査読あり, その他, 情報処理学会論文誌数理モデル化と応用, 仮想風景におけるVR酔い低減のための一手法, 笹山 琴由; 五味田 啓; 加藤 嘉明; 城 和貴, 2021年03月, 14, 2, 13, 23
査読無し, その他, 情報処理学会数理モデル化と問題解決研究会, レーザー打音分析によるコンクリート内部状態の推定, 千代延 未帆; 城 和貴; 石川 由羽; 錦野 将元, 2021年02月, 132, 4, 1, 6
査読無し, その他, 情報処理学会数理モデル化と問題解決研究会, 近代書籍における文字切り出し手法の検討, 飯田 紗也香; 竹本 有紀; 石川 由羽; 髙田 雅美; 城 和貴, 2021年02月, 132, 4, 1, 6
査読無し, その他, 情報処理学会数理モデル化と問題解決研究会, 速度低下錯視を用いた VR 酔いの低減手法, 笹山琴由; 五味田啓; 加藤嘉明; 城和貴, 2020年12月, 127, 11, 1, 6
査読無し, その他, 情報処理学会数理モデル化と問題解決研究会, GPを用いたJAZZ実時間自動生成システムの試作, 井上 彩; 城 和貴, 2020年12月, 131, 18, 1, 6
査読無し, その他, 情報処理学会数理モデル化と問題解決研究会, 邦字新聞における近代文語体と現代口語体の自動翻訳の検討, 稲見 郁乃; 竹本 有紀; 石川 由羽; 高田 雅美; 上田 薫; 城 和貴, 2020年12月, 131, 12, 1, 6
査読無し, その他, 情報処理学会数理モデル化と問題解決研究会, 地震予知を目的とした DEMETER 衛星データの雷ノイズ除去, 石通彩加; 竹本有紀; 石川由羽; 高田雅美; 城和貴, 2020年02月, 127, 24, 1, 6
査読あり, 英語, Proceedings of the 1st International ACM Conference on Multimedia Information Retrieval, MIR2008, Co-located with the 2008 ACM International Conference on Multimedia, MM'08, Searching musical audio datasets by a batch of multi-variant tracks, Yi Yu; J. Stephen Downie; Lei Chen; Vincent Oria; Kazuki Joe, Multi-variant music tracks are those audio tracks of a particular song which are sung and recorded by different people (i.e., cover songs). As music social clubs grow on the Internet, more and more people like to upload music recordings onto such music social sites to share their own homeproduced albums and participate in Internet singing contests. Therefore it is very important to explore a computerassisted evaluation tool to detect these audio-based multivariant tracks. In this paper we investigate such a task: the original track of a song is embedded in datasets, with a batch of multi-variant audio tracks of this song as input, our retrieval system returns an ordered list by similarity and indicates the position of relevant audio track. To help process multi-variant audio tracks, we suggest a semantic indexing framework and propose the Federated Features (FF) scheme to generate the semantic summarization of audio feature sequences. The conjunction of federated features with three typical similarity searching schemes, K-Nearest Neighbor (KNN), Locality Sensitive Hashing (LSH), and Exact Euclidian LSH (E2LSH), is evaluated. From these findings, a computer-assisted evaluation tool for searching multi-variant audio tracks was developed to search over large musical audio datasets. Copyright 2008 ACM., 2008年, 121, 127, 研究論文(国際会議プロシーディングス), 10.1145/1460096.1460117
査読あり, 英語, MM'08 - Proceedings of the 2008 ACM International Conference on Multimedia, with co-located Symposium and Workshops, COSIN: Content-based retrieval system for cover songs, Yi Yu; J. Stephen Downie; Fabian Moerchen; Lei Chen; Kazuki Joe; Vincent Oria, We develop a content-based audio COver Song IdeNtification (COSIN) system to detect/group cover songs.The COSIN takes music audio content as input and performs similarity searching to locate variants of the input (i.e., cover versions). Identified cover songs are returned in the rank order according to their similarity to the input.The COSIN also incorporates a set of tools to evaluate retrieval performance so researchers can explore different retrieval schemes and parameters (e.g. recall, precision).The COSIN utilizes a suite of techniques to detect cover songs including: Pitch + Dynamic Programming (DP), Chroma + DP, and Semantic Feature Summarization (SFS) + Hash-Based Approximate Matching (HBAM). Demonstration system shows that COSIN is a very potential music content retrieval tool. Running some music retrieval schemes on COSIN platform, recent experiments with SFS + LSH Variants demonstrate a nicely balanced efficiency (search speed) v. performance (search accuracy) tradeoff., 2008年, 987, 988, 研究論文(国際会議プロシーディングス), 10.1145/1459359.1459542
査読無し, 日本語, 情報処理学会数理モデル化と問題解決研究会, 近代書籍における低出現頻度文字種の獲得, 藤田未希; 竹本有紀; 石川由羽; 髙田雅美; 城和貴, 2019年12月, 126, 6, 1, 6, 研究論文(研究会,シンポジウム資料等)
査読無し, その他, 情報処理学会数理モデル化と問題解決研究会, ホワイトボードからの文字抽出手法の検討, 山本祐美; 本郷仁志; 森吉弘; 城和貴, 2019年12月, 126, 14, 1, 6, 研究論文(研究会,シンポジウム資料等)
査読あり, 英語, The 2019 International Conference on Parallel and Distributed Processing Techniques and Applications, Applying CNNs to Early-Modern Printed Japanese Character Recognition, Suzuka Yasunami; Norie Koiso; Yuki Takemoto; Yu Ishikawa; Masami Takata; Kazuki Joe, 2019年08月, 189, 195, 研究論文(国際会議プロシーディングス)
査読あり, その他, The 2019 International Conference on Parallel and Distributed Processing Techniques and Applications, Structure of Neural Network Automatically Generating Fonts for Early-Modern Japanese Printed Books, Yuki Takemoto; Yu Ishikawa; Masami Takata; Kazuki Joe, 2019年08月, 182, 188, 研究論文(国際会議プロシーディングス)
査読あり, 英語, The 2019 International Conference on Parallel and Distributed Processing Techniques and Applications, Layout Analysis using Semantic Segmentation for Imperial Meeting Minutes, Sayaka Iida; Yuki Takemoto; Yu Ishikawa; Masami Takata; Kazuki Joe, 2019年08月, 135, 141, 研究論文(国際会議プロシーディングス)
査読あり, 英語, The 2018 International Conference on Parallel and Distributed Processing Techniques and Applications, Final Edition, Optimization of Auto Mesh Generation Using Genetic Algorithm, 城和貴, 2018年, 354-360
査読あり, 英語, The 2018 International Conference on Parallel and Distributed Processing Techniques and Applications, Final Edition, Automatic Font Generation for Early-Modern Japanese Printed Books, 城和貴, 2018年, 326-332
査読あり, 英語, The 2018 International Conference on Parallel and Distributed Processing Techniques and Applications, Final Edition, Refinement of a real-time super-resolution FPGA circuit, 城和貴, 2018年, 347-353
査読あり, 英語, 2017 14th International Conference on Ubiquitous Robots and Ambient Intelligence, URAI 2017, Estimating emotion of user via communicative stuffed-toy device with pressure sensors using fuzzy reasoning, Tomoko Yonezawa; Haruka Mase; Hirotake Yamazoe; Kazuki Joe, In this research, we aim to perform emotion estimation based on the touching manner of communication using our stuffed-toy device that we had previously proposed in order to reduce the user's stress. In this paper, we propose a method to classify a user's emotions into three types: comfort, discomfort, and unknown, by adopting a fuzzy reasoning using histogram distribution of peaks of pressure sensor values. The results of the experiment confirmed the accuracy of classification as 69.1% of the correct answer rate when the results estimated as unknown are included in the correct answers, and 63.1% of the correct answer rate when the unknown results are excluded from the correct answer. We also discuss improvements of future methods., 2017年07月25日, 916, 921, 研究論文(国際会議プロシーディングス), 10.1109/URAI.2017.7992865
査読あり, 英語, The 2017 International Conference on Parallel and Distributed Processing Techniques and Applications, Final Edition, Hand Shape Recognition using the sEMG of the Upper Extremity, 城和貴, 2017年, 295-301
査読あり, 英語, The 2017 International Conference on Parallel and Distributed Processing Techniques and Applications, Final Edition, Validation of EEG Authentication Accuracy with Electrode Slippage, 城和貴, 2017年, 302-308
査読あり, 日本語, 情報処理学会論文誌数理モデル化と応用, 脳波から求めた複数特徴量の組み合わせによる生体認証, 城和貴; 石川由羽; 高田雅美, 2017年, 10, 1, 22-32
査読あり, 英語, The 2017 International Conference on Parallel and Distributed Processing Techniques and Applications, Final Edition, Early-Modern Printed Character Recognition using Ensemble Learning, 城和貴, 2017年, 288-294
査読あり, 英語, The 2016 International Conference on Parallel and Distributed Processing Techniques and Applications, Final Edition, Correlation of proximity voluntary muscles EMG and EEG, 城和貴, 2016年, 401-407
査読あり, 英語, The 2016 International Conference on Parallel and Distributed Processing Techniques and Applications, Final Edition, Real-Time Super Resolution: FPGA Implementation for the ICBI Algorithm, 城和貴, 2016年, 415-420
査読あり, 日本語, 情報処理学会論文誌数理モデル化と応用, 近代書籍を対象とした多フォント漢字認識, 城和貴; 粟津 妙華; 上坂 和美; 高田 雅美, 2016年, 9, 2, 33-40
査読あり, 英語, The 2016 International Conference on Parallel and Distributed Processing Techniques and Applications, Final Edition, Biometric Authentication based on Multi-feature Combination using EEG, 城和貴, 2016年, 401-407
査読あり, 英語, The 2016 International Conference on Parallel and Distributed Processing Techniques and Applications, Final Edition, Comparison of Feature Extraction Methods for Early-Modern Japanese Printed Character Recognition, 城和貴, 2016年, 408-414
査読あり, 英語, The 2015 International Conference on Parallel and Distributed Processing Techniques and Applications, A Music Composition Model with Genetic Programming -A Case Study of Chord Progression and Bassline-, 城和貴; Kanae Kunimatsu; Yu Ishikawa; Masami Takata, 2015年07月, 2, 256-262
査読あり, 英語, The 2015 International Conference on Parallel and Distributed Processing Techniques and Applications, Sign Language Recognition using Leap Motion Controller, 城和貴; Makiko Funasaka; Yu Ishikawa; Masami Takata, 2015年07月, 2, 263-269
査読あり, 英語, The 2015 International Conference on Parallel and Distributed Processing Techniques and Applications, An Effective and Interactive Training Data Collection Method for Early-Modern Japanese Printed Character Recognition, 城和貴; Kazumi Kosaka; Taeka Awazu; Yu Ishikawa; Masami Takata, 2015年07月, 2, 276-282
査読あり, 英語, The 2015 International Conference on Parallel and Distributed Processing Techniques and Applications, Visualization of Sensory Weight for Shouldering Randseru, 城和貴; Hitomi Oigawa; Yu Ishikawa; Masami Takata, 2015年07月, 2, 297-303
査読あり, 英語, The 2015 International Conference on Parallel and Distributed Processing Techniques and Applications, Evaluating Elements of Communicative Stuffed-toy Device Describes Scripts on SNS, 城和貴; Haruka Mase; Tomoko Yonezawa, 2015年07月, 2, 310-316
査読あり, 英語, The 2015 International Conference on Parallel and Distributed Processing Techniques and Applications, A Personal Classification Method Using Spatial Information of Multi-channel EEG, 城和貴; Yu Ishikawa; Chinami Yoshida; Masami Takata; Hiroyasu Kamo, 2015年07月, 2, 229-235
査読あり, 日本語, 情報処理学会論文誌数理モデル化と応用, 活字データの分類を用いた進化計算による近代書籍からのルビ除去, 城和貴; 粟津 妙華; 高田 雅美, 2015年03月, 8, 1, 72-79
査読あり, 英語, The 2014 International Conference on Parallel and Distributed Processing Techniques and Applications, Validation of EEG Personal Authentication with Multi-channels and Multi-tasks, 城和貴; Yu Ishikawa; Chinami Yoshida; Masami Takata, 2014年07月, 2, 182-188
査読あり, 英語, 3rd International Conference on Pattern Recognition Applications and Methods, A Multi-Fonts Kanji Character Recognition Method for Early-Modern Japanese Printed Books with Ruby Characters, 城和貴, 2014年03月, 637-645
査読あり, 英語, 7th International Joint Conference on Biomedical Engineering Systems and Technologies, Special Session on Multivariable Processing for Biometric Systems, Biometrics Authentication using Another Feature of Heartbeat Waveform, 城和貴, 2014年03月, 311-317
査読あり, 英語, SEMANTIC TECHNOLOGY, Ontology Construction Support for Specialized Books, Yuki Eguchi; Yuri Iwakata; Minami Kawasaki; Masami Takata; Kazuki Joe, In this paper, we present a support system for ontology construction just based on a given specialized book with lowest possible cost. The system tries to combine minimum hand and mostly automatic constructions. It extracts required information for the ontology design from the specialized book with presenting yes-no selections to an expert of the specialized book. The ontology construction is performed automatically just using the answers of the expert. The constructed ontology is reliably and highly technical since it is constructed on the basis of the specialized book. In addition, since user operations are restricted only to yes-no selection, any expert can make use of our system without any special knowledge about ontology., 2014年, 8388, 159, 174, 研究論文(国際会議プロシーディングス), 10.1007/978-3-319-06826-8_13
査読あり, 英語, 2014 7TH BIOMEDICAL ENGINEERING INTERNATIONAL CONFERENCE (BMEICON), Biometrics Authentication Based on Chaotic Heartbeat waveform, Ryoko Nomura; Yu Ishikawa; Tomohiro Umeda; Masami Takata; Hiroyasu Kamo; Kazuki Joe, In this paper, we propose a personal authentication method using heartbeat waveforms to enhance the security in wireless communication. In the heartbeat waveforms of a human being, it is known that there are discriminative characteristic features and the mimicking is very difficult. Therefore, their application to personal authentication has been studied. Existing method performs personal authentication using a chaos indicator of ECG. In this paper, we propose an authentication method that extends the index in order to increase the accuracy of authentication. To validate that our personal authentication method is applicable, we perform some experiments to show that our method provides better authentication than existing one., 2014年, 1-5, 研究論文(国際会議プロシーディングス)
査読あり, 日本語, 情報処理学会論文誌数理モデル化と応用, 遺伝的プログラミングを用いた近代書籍からのルビ除去, 城和貴; 粟津妙華; 高田雅美, 2013年, 6, 2, 53-62
査読あり, 英語, The 2013 International Conference on Parallel and Distributed Processing Techniques and Applications, Locality Analysis for Characterizing Applications Based on Sparse Matrices, 城和貴; Noboru Tanabe; Sonoko Tomimori; Masami Takata, 2013年, 2, 299-315
査読あり, 日本語, 情報処理学会論文誌数理モデル化と応用, 古典油彩画グレーズ技法シミュレーションの高速化, 城和貴; 篠澤 和恵; 楜沢 順; 高田 雅美, 2013年, 6, 1, 119-127
査読あり, 英語, 5th International Conference on Future Computational Technologies and Applications, Future Irregular Computing with Memory Accelerators, 城和貴, 2013年, 74-80
査読あり, 日本語, 情報処理学会論文誌数理モデル化と応用, 視覚障害者のための墨字学習支援システムの開発と評価, 城和貴; 松本多恵; 伊奈諭; 高田雅美, 2013年, 6, 2, 128-136
査読あり, 英語, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Character of graph analysis workloads and recommended solutions on future parallel systems, Noboru Tanabe; Sonoko Tomimori; Masami Takata; Kazuki Joe, Graph500 is a benchmark suite for big data analysis. Matrices used for Graph500 inherit the properties of graph analysis such as breadth first search for SNS and PageRank for web searching engine. Especially power saving is very important for its execution on future massively parallel processors and clouds. The spatial locality of sparse matrices used for Graph500 and its behaviors on cache memory are investigated. The experimental results show the spatial locality of sparse matrices used for Graph500 is very low. It is very difficult to solve the problem by just software approach because of the huge size and the randomness of their accesses. Therefore, we recommend hardwired scatter/gather functions at memory side. They improve the processing speed in an order of magnitude. For achieving both of low power and high throughput of random access, we recommend implementing hardwired scatter/gather functions on logic-base in Hybrid Memory Cube (HMC). We also describe brief considerations of the power saving in the case of low cache hit rate application such as graph500. For example, when the hit rate is 15%, the power saving ratio of memory access is about 30-fold. © Springer International Publishing Switzerland 2013., 2013年, 8285, 1, 402, 415, 研究論文(国際会議プロシーディングス), 10.1007/978-3-319-03859-9_35
査読あり, 日本語, 情報処理学会論文誌数理モデル化と応用, α波構成要素の切出しと解析, 城和貴; 石川由羽; 高田雅美, 2013年, 6, 2, 93-103
査読あり, 英語, The 2012 International Conference on Parallel and Distributed Processing Techniques and Applications, AR Based Spatial Reasoning Capacity Training for Students, 城和貴, 2012年07月, II, 750-755
査読あり, 英語, The 2012 International Conference on Parallel and Distributed Processing Techniques and Applications, Real Time Spatiotemporal Biological Stress Level Checking, 城和貴, 2012年07月, II, 744-749
査読あり, 英語, ASIASIM 2012, PT I, Pre-motion Based VR Object Manipulation Definitions and Preliminary Experiments, Shiori Mizuno; Asuka Komeda; Naoko Yoshii; Tomoko Yonezawa; Masami Takata; Kazuki Joe, In this paper, we propose a new method to manipulate objects by using user's pre-motion in a VR environment. Users are provided with an instinctually easy interface to VR with using their natural behaviors as their commands for manipulating VR objects. Hand pre-shaping is known as human's unconscious behavior to face a small object to be grasped. The definitions of pre-motions are determined by users' initial behaviors. Therefore, we present the definitions to classify characteristics of the kinds of objects and pre-motions. Using the above definitions, we develop a prototype system to validate the classifications. Consequently, manipulating VR objects by pre-motion is possible and promising., 2012年, 323, 499, +, 研究論文(国際会議プロシーディングス)
査読あり, 英語, 5TH BIOMEDICAL ENGINEERING INTERNATIONAL CONFERENCE (BMEICON 2012), Constitution and Phase Analysis of Alpha Waves, Yu Ishikawa; Masami Takata; Kazuki Joe, Alpha waves are electroencephalogram discovered by Hans Berger in 1929 and have been studied by many researchers. Recently the amplitude and the phase of alpha waves attract attention, and various studies are reported. So far the alpha waves are not concretely modeled to explain how they are constituted. In this paper, we analyze the phase of alpha waves for their future modeling. As pre-processing, we extract alpha wave data from EEGs and divide the data into alpha wave constituent waves. Analyzing the feature quantity of the alpha wave constituent waves, it turns out that the amplitude and the wavelength of alpha wave constituent waves are correlated with a constant time delay., 2012年, 研究論文(国際会議プロシーディングス)
査読あり, 日本語, 情報処理学会論文誌コンピューティングシステム, 長行を折畳む疎行列ベクトル積方式とGather機能付メモリによる高速化, 城和貴; 田邊昇; 小郷絢子; 小川裕佳; 高田雅美, 2012年, 5, 4, 112-124
査読あり, 英語, ASIASIM 2012, PT I, A Multi-resolution Display Method for Digital Archives of Cultural Properties Using Polygon Reconsruction, Megumi Okumoto; Yuri Iwakata; Asuka Komeda; Tomoko Yonezawa; Masami Takata; Kazuki Joe, In this paper, we propose a multi-resolution display method for digital archives of cultural properties using polygon reconstruction. When digital archives of cultural properties are displayed in VR systems for research purpose, mesh resolution had to be changeable on the demand of user. Additionally, for user's comfortability, polygon reconstruction needs to be executed in real-time. History of preliminary polygon reductions is adopted for the proposed method so that polygon reconstruction is executed in real-time. To validate the VR system using the polygon reconstruction based multi-resolution display method, some experiments are performed. As the result of the experiments, the proposed method is performed within 1.0 second, which is considered as "real-time" by the definition of typical user-interface., 2012年, 323, 190, +, 研究論文(国際会議プロシーディングス)
査読あり, 日本語, NVIDIA GTC Workshop Japan 2011, GPGPU向けメモリアクセラレータ, 田邊昇; Boonyasitpichai Nuttapon; 中條拓伯; 小郷絢子; 高田雅美; 城和貴, 2011年07月, 研究論文(学術雑誌)
査読あり, 英語, The 2011 International Conference on Parallel and Distributed Processing Techniques and Applications, A Real-time Analysis Environment for a Wireless BMI Device Enobio, 城和貴, 2011年07月, II, 739-744
査読あり, 英語, The 2011 International Conference on Parallel and Distributed Processing Techniques and Applications, Evaluation of the SVM Based Multi-Fonts Kanji Character Recognition Method for Early-Modern Japanese Printed Books, 城和貴, 2011年07月, II, 727-732
査読あり, 英語, The 2011 International Conference on Parallel and Distributed Processing Techniques and Applications, Optimization of the Particle-based Volume Rendering for GPUs with Hiding Data Transfer Latency, 城和貴, 2011年07月, II, 733-738
査読あり, 英語, PROCEEDINGS OF THE FIRST WORKSHOP ON IRREGULAR APPLICATIONS: ARCHITECTURES AND ALGORITHM (IAAA'11), A Memory Accelerator with Gather Functions for Bandwidth-bound Irregular Applications, Noboru Tanabe; Boonyasitpichai Nuttapon; Hironori Nakajo; Yuka Ogawa; Junko Kogou; Masami Takata; Kazuki Joe, Compute intensive processing can be easily accelerated using processors with many cores such as GPUs. However, memory bandwidth limitation becomes serious year by year for memory bandwidth intensive applications such as sparse matrix vector multiplications (SpMV). In order to accelerate memory bandwidth intensive applications, we have proposed a memory system with additional functions of scattering and gathering. For the preliminary evaluation of our proposed system, we assumed that the throughput of the memory system was sufficient. In this paper, we propose a memory system with scattering and gathering using many narrow memory channels. We evaluate the feasible throughput of the proposed memory system based on DDR3 DRAM with the modified DRAMsim2 simulator. In addition, we evaluate the performance of SpMV using our method for the proposed memory system and a GPU. We have confirmed the proposed memory system has good performance and good stability for matrix shape variation using fewer pins for external memory., 2011年, 35, 42, 研究論文(国際会議プロシーディングス)
査読あり, 英語, PROCEEDINGS OF THE 19TH INTERNATIONAL EUROMICRO CONFERENCE ON PARALLEL, DISTRIBUTED, AND NETWORK-BASED PROCESSING, Scaleable Sparse Matrix-Vector Multiplication with Functional Memory and GPUs, Noboru Tanabe; Yuuka Ogawa; Masami Takata; Kazuki Joe, Sparse matrix-vector multiplication on GPUs faces to a serious problem when the vector length is too large to be stored in GPU's device memory. To solve this problem, we propose a novel software-hardware hybrid method for a heterogeneous system with GPUs and functional memory modules connected by PCI express. The functional memory contains huge capacity of memory and provides scatter/gather operations. We perform some preliminary evaluation for the proposed method with using a sparse matrix benchmark collection. We observe that the proposed method for a GPU with converting indirect references to direct references without exhausting GPU's cache memory achieves 4.1 times speedup compared with conventional methods. The proposed method intrinsically has high scalability of the number of GPUs because intercommunication among GPUs is completely eliminated. Therefore we estimate the performance of our proposed method would be expressed as the single GPU execution performance, which may be suppressed by the burst-transfer bandwidth of PCI express, multiplied with the number of GPUs., 2011年, 101, 108, 研究論文(国際会議プロシーディングス), 10.1109/PDP.2011.92
査読あり, 英語, CHEMICAL PHYSICS LETTERS, Analysis of motion features for molecular dynamics simulation of proteins, Mayumi Kamada; Mikito Toda; Masakazu Sekijima; Masami Takata; Kazuki Joe, Recently, a new method for time series analysis using the wavelet transformation has been proposed by Sakurai et al. We apply it to molecular dynamics simulation of Thermomyces lanuginosa lipase (TLL). Introducing indexes to characterize collective motion of the protein, we have obtained the following two results. First, time evolution of the collective motion involves not only the dynamics within a single potential well but also takes place wandering around multiple conformations. Second, correlation of the collective motion between secondary structures shows that collective motion exists involving multiple secondary structures. We discuss future prospects of our study involving 'disordered proteins'. (C) 2010 Elsevier B.V. All rights reserved., 2011年01月, 502, 4-6, 241, 247, 研究論文(学術雑誌), 10.1016/j.cplett.2010.12.028
査読あり, 日本語, 情報処理学会論文誌数理モデル化と応用, 電車ノイズを含む地電位差データからの矩形状地震前駆的シグナル自動抽, 石川千里; 豊島良美; 高田雅美; 長尾年恭; 城和貴, 2010年, 3, 2, 22-31
査読あり, 英語, The 2010 International Conference on Parallel and Distributed Processing Techniques and Applications, Optimization of GROMACS for CUDA Environments, 城和貴, 2010年, II, 679-685
査読あり, 英語, The 2010 International Conference on Parallel and\nDistributed Processing Techniques and Applications, Feature Detection of Volume Data Based on Topology and its Application to Database Construction, 城和貴, 2010年, II, 665-671
査読あり, 英語, The 2010 International Conference on Parallel and Distributed Processing Techniques and Applications, A Music Composition Model with Genetic Programming, 城和貴, 2010年, II, 686-692
査読あり, 英語, The 2010 International Conference on Parallel and\nDistributed Processing Techniques and Applications,, Multi-Regression Analysis of Music Impressions for\nMusic Evaluation, 城和貴, 2010年, II, 693-702
査読あり, 英語, The 2010 International Conference on Parallel and Distributed Processing Techniques and Applications, Update Dependence Analysis of Web Pages for Incremental Crawling, 城和貴, 2010年, II, 672-678
査読あり, 英語, The 2009 International Conference on Parallel and Distributed Processing Techniques and Applications, The Architecture of Visualization System using Memory with Memory-Side Gathering and CPUs with DMA-type Memory Accessing, 城和貴, 2009年, II, 427-433
査読あり, 英語, The 2009 International Conference on Parallel and Distributed Processing Techniques and Applications, Modeling the Developing Process of Depth Perception for Elementary Students, 城和貴, 2009年, II, 735-741
査読あり, 英語, The 2009 International Conference on Parallel and Distributed Processing Techniques and Applications, Recognition of Multi-Fonts Character in Early-Modern Printed Books, 城和貴, 2009年, II, 728-734
査読あり, 日本語, 情報処理学会論文誌数理モデル化と応用, Iアプリによる数値計算実アプリケーションの実装例, 程 暁紅; 沼野 なぎさ; 高田 雅美; 城 和貴, 2009年, 2, 2, 38-46
査読あり, 英語, The 2009 International Conference on Parallel and Distributed Processing Techniques and Applications, Clustering the Temporal Sequences Feature of 3D Protain Structure, 城和貴, 2009年, II, 749-755
査読あり, 英語, The 2009 International Conference on Parallel and Distributed Processing Techniques and Applications, Optimization of a GP Application for the Cell Processor, 城和貴, 2009年, II, 742-748
査読あり, 英語, The 2009 International Conference on Parallel and Distributed Processing Techniques and Applications, A Deterioration Lawns Model Constructed from Image Data, 城和貴, 2009年, II, 721-727
査読あり, 英語, The 2009 International Conference on Parallel and Distributed Processing Techniques and Applications, The Architecture of Visualization System using Memory with Memory-Side Gathering and CPUs with DMA-type Memory Accessing, 城和貴, 2008年07月, 2, 427-433
査読あり, 英語, IEICE TRANSACTIONS ON INFORMATION AND SYSTEMS, Efficient query-by-content audio retrieval by Locality Sensitive Hashing and partial sequence comparison, Yi Yu; Kazuki Joe; J. Stephen Downie, This paper investigates suitable indexing techniques to enable efficient content-based audio retrieval in large acoustic databases. To make an index-based retrieval mechanism applicable to audio content, we investigate the design of Locality Sensitive Hashing (LSH) and the partial sequence comparison. We propose a fast and efficient audio retrieval framework of query-by-content and develop an audio retrieval system. Based on this framework, four different audio retrieval schemes, LSH-Dynamic Programming (DP), LSH-Sparse DP (SDP), Exact Euclidian LSH (E(2)LSH)DP, E(2)LSH-SDP, are introduced and evaluated in order to better understand the performance of audio retrieval algorithms. The experimental results indicate that compared with the traditional DP and the other three compititive schemes, E2LSH-SDP exhibits the best tradeoff in terms of the response time, retrieval accuracy and computation cost., 2008年06月, E91D, 6, 1730, 1739, 研究論文(学術雑誌), 10.1093/ietisy/e91-d.6.1730
査読あり, 英語, ISM: 2008 IEEE INTERNATIONAL SYMPOSIUM ON MULTIMEDIA, Using Exact Locality Sensitive Mapping to Group and Detect Audio-Based Cover Songs, Yi Yu; J. Stephen Downie; Fabian Moerchen; Lei Chen; Kazuki Joe, Cover song detection is becoming a very hot research topic when plentiful personal music recordings or performance are released on the Internet. A nice cover song recognizer helps us group and detect cover songs to improve the searching experience. The traditional detection is to match two musical audio sequences by exhaustive pairwise comparisons. Different from the existing work, our aim is to generate a group of concatenated feature sets based on regression modeling and arrange them by indexing-based approximate techniques to avoid complicated audio sequence comparisons. We mainly focus on using Exact Locality Sensitive Mapping (ELSM) to join the concatenated feature sets and soft hash values. Similarity-invariance among audio sequence comparison is applied to define an optimal combination of several audio features. Soft hash values are pre-calculated to help locate searching range more accurately. Furthermore, we implement our algorithms in analyzing the real audio cover songs and grouping and detecting a batch of relevant cover songs embedded in large audio datasets., 2008年, 302, +, 研究論文(国際会議プロシーディングス)
査読あり, 英語, The 2008 International Conference on Parallel and Distributed Processing Techniques and Applications, Preliminary Modeling of Yeast with Alcohol Tolerant Property for the E-Cell System, 城和貴, 2008年, II, 776-782
査読あり, 英語, The 2008 International Conference on Parallel and Distributed Processing Techniques and Applications, Real-time 3D Movie Generation by Anaglyph for Live Streaming, 城和貴, 2008年, II, 758-763
査読あり, 英語, HIGH-PERFORMANCE COMPUTING, Development of an interactive visual data mining system for atmospheric science, Chiemi Watanabe; Eriko Touma; Kazuko Yamauchi; Katsuyuki Noguchi; Sachiko Hayashida; Kazuki Joe, In atmospheric science, 3D visualization techniques have been mainly used to create impressive presentation in recent decades. However, from the viewpoint of utilize for visual data mining, 3D visualization methodology has difficulties in becoming wide spread because most conventional and established way is to make 2D diagrams consisting of two dimensions of a temporal transitional 3D grid. From these observations, we have been developing a quick look tool of atmospheric science data for 3d visual data mining. We expect that scientists can utilize this tool for finding out 2D diagrams from the data by using various 2D or 3D visualization methods, and become accustomed themselves to 3D visualization methods., 2008年, 4759, 279, +, 研究論文(国際会議プロシーディングス)
査読あり, 英語, GRAPP 2008: PROCEEDINGS OF THE THIRD INTERNATIONAL CONFERENCE ON COMPUTER GRAPHICS THEORY AND APPLICATIONS, An interactive 3D visualization model by live streaming for remote scientific visualization, Eriko Touma; Satomi Hara; Mari Kurumi; Yuri Shirakawa; Chisato Ishikawa; Masami Takata; Takeshi Horinouchi; Kazuki Joe, Recent improvement of high-end GPUs has made it possible to perform real-time 3D visualization such as volume rendering and 3D contour plot for scientific data locally. A web browser based remote 3D visualization by visualization servers is attractive, but data transfer overhead prevents from performing interactive operations. We propose an interactive remote 3D visualization model by live streaming for geophysical fluids research. In this model, we use live streaming flash media for the web browser based operations keeping minimum quality of data analysis and minimum bit rate for live streaming of flash media. Preliminary experiments with a prototype system validate the effectiveness of our proposing model., 2008年, 409, 414, 研究論文(国際会議プロシーディングス)
査読あり, 英語, The 2008 International Conference on Parallel and Distributed Processing Techniques and Applications, A Room Modeling Method for 3D Video, 城和貴, 2008年, II, 751-757
査読あり, 英語, ISM WORKSHOPS 2007: NINTH IEEE INTERNATIONAL SYMPOSIUM ON MULTIMEDIA - WORKSHOPS, PROCEEDINGS, An evaluation of feature extraction for query-by-content audio information retrieval, Yi Yu; J. Stephen Downie; Kazuki Joe, Content-based audio information retrieval is one of the most interesting and fast-growing research areas. Suitable feature sets can help to reduce the tedious computation time and speed up retrieval. In this paper we report a study of the music spectral properties aimed at the acoustic-based music data similarity measurement and show that the spectral features of adjacent flames are highly correlated. Based on such a case study we mainly focus on making an evaluation of feature choice in the three aspects: storage, computation and retrieval ratio. The extensive evaluations confirm the effectiveness of feature merge in quickening sequence matching for query-by-content audio retrieval and show that MFCC with feature merge is the best tradeoff among storage requirement, computation cost and retrieval ratio., 2007年, 297, +, 研究論文(国際会議プロシーディングス), 10.1109/ISM.Workshops.2007.57
査読あり, 日本語, 情報処理学会論文誌:数理モデル化と応用, アブストラクトを用いた原子分子物理学分野の論文分類支援システムの設計と実装, 城和貴; 柏木裕恵; 高田雅美; 佐々木明, 2007年, 48, SIG15, 189-198
査読あり, 英語, The 2007 International Conference on Parallel and Distributed Processing Techniques and Applications, Construction of a Paper Classification System Using SVM, 城和貴, 2007年, II, 759-765
査読あり, 英語, The 2007 International Conference on Parallel and Distributed Processing Techniques and Applications, Development of Molecular Dynamics Simulation based Flexible Docking System, 城和貴, 2007年, II, 739-745
査読あり, 英語, ADVANCES IN MULTIMEDIA MODELING, PT 1, Similarity searching techniques in content-based audio retrieval via hashing, Yi Yu; Masami Takata; Kazuki Joe, With this work we study suitable indexing techniques to support efficient, content-based music retrieval in large acoustic databases. To obtain the index-based retrieval mechanism applicable to audio content, we pay the most attention to the design of Locality Sensitive Hashing (LSH) and the partial sequence comparison, and propose a fast and efficient audio retrieval framework of query-by-content. On the basis of this indexable framework, four different retrieval schemes, LSH-Dynamic Programming (DP), LSH-Sparse DP (SDP), Exact Euclidian LSH ((ELSH)-L-2)-DP, (ELSH)-L-2-SDP, are presented and estimated in order to achieve an extensive understanding of retrieval algorithms performance. The experiment results indicate that compared to other three schemes, (ELSH)-L-2-SDP exhibits best tradeoff in terms of the response time, retrieval ratio, and computation cost., 2007年, 4351, LNCS4351, 397, 407, 研究論文(国際会議プロシーディングス)
査読あり, 英語, The 2007 International Conference on Parallel and Distributed Processing Techniques and Applications, Auto Classification of Feces for Health Condition Analysis, 城和貴, 2007年, II, 979-985
査読あり, 英語, The 2007 International Conference on Parallel and Distributed Processing Techniques and Applications, Optimization of Molecular Dynamics Simulation on Cell Processor, 城和貴, 2007年, II, 780-786
査読あり, 英語, The 2007 International Conference on Parallel and Distributed Processing Techniques and Applications, Dvelopment of SVM based Prediction System for Metalbinding Sites in Protein,, 城和貴, 2007年, II, 972-978
査読あり, 英語, The 2007 International Conference on Parallel and Distributed Processing Techniques and Applications, 3D Visualization System Gateau for Atmospheric Science: Design Concept and Practical Evaluation, 城和貴, 2007年, II, 766-772
査読あり, 日本語, 情報処理学会論文誌数理モデル化と応用, 情報爆発時代に向けた大気科学分野のための3次元可視化モデルの提案, 当麻英梨子; 高田雅美; 城和貴, 2007年, 48, SIG19, 86-97
査読あり, 英語, 1st Workshop on Learning the Semantics of Audio Signals, Index-based Similarity Searching with Partial Sequence Comparison for Query-by-Content Audio Retrieval, 城和貴, 2006年12月, 76-86
査読あり, 英語, 2nd IEEE International Conference on e-Science and Grid Computing, Automatic Viewpoint Selection for a Visualization I/F in a PSE, 城和貴, ISBN0-7695-2734-5, 2006年12月, CD-ROM
査読あり, 日本語, 情報処理学会論文誌:数理モデル化と応用, 可視化されたデータ依存の分類とその並列化手法への適用, 城和貴; 山口智美; 笹倉万里子, 2006年, 47, SIG14, 179-186
査読あり, 英語, 1st Workshop on Learning the Semantics of Audio Signals, Index-based Similarity Searching with Partial Sequence Comparison for Query-by-Content Audio Retrieval, 城和貴, 2006年, 76-86
査読あり, 英語, The 2006 International Conference on Parallel and Distributed Processing Techniques and Applications, Design and Implementation of Gateau API for HDF-EOS, 城和貴, 2006年, 867-873
査読あり, 英語, The 2006 International Conference on Parallel and Distributed Processing Techniques and Applications, A 3D Visualization System for Intuitive Analysis of Atmospheric Science, 城和貴, 2006年, 860-866
査読あり, 英語, The 2006 International Conference on Parallel and Distributed Processing Techniques and Applications, Design of an Integrated Database System for Short-Term Earthquake Prediction, 城和貴, 2006年, 853-859
査読あり, 英語, The 2006 International Conference on Parallel and Distributed Processing Techniques and Applications, Another RSS Reader with Visualization of Structured Documents and Effective Navigation, 城和貴, 2006年, 874-880
査読あり, 英語, Proceedings of the 8th Joint Conference on Information Sciences, Vols 1-3, Rule-set evaluation of article relevance for autonomous bibliography databases, L Pichl; M Suzuki; D Kato; K Joe; A Sasaki, Text relevance assessment in data mining is usually dealt with by various artificial intelligence based methods. This paper focuses on one important case - computerized decision whether a certain scientific article contains numerical data of interest for a specialized research database or not. The area of applications covers activities of most major research data centers, with applications ranging from proteomics to fusion plasma research. As a part of coordinated research on plasma-process bibliography databases at the National Institute for Fusion Science in Japan (NIFS), we have recently developed a linux system that automates the process of data collection, data extraction and database input, with a customizable interface to relevance-assessment software modules. Here we deal with the article relevance assessment from two major viewpoints: a rule-based decision making system based on the analysis of figure and table caption texts, and a machine learning system that analyzes html abstracts of the articles. It is demonstrated that a combination of the two methods may result in a highly specific and sensitive relevance assessment system. Features unique in atomic and molecular data mining are also discussed., 2005年, 1453, 1456, 研究論文(国際会議プロシーディングス)
査読あり, 英語, 6th International Conference on Music Information Retrieval, Towards A Fast and Efficient Match Algorithm for Content-Based Music Retrieval on Acoustic Data, 城和貴, 2005年, 696-701
査読あり, 英語, DATABASES IN NETWORKED INFORMATION SYSTEMS, PROCEEDINGS, Networked mining of atomic and molecular data from electronic journal databases on the Internet, L Pichl; M Suzuki; K Joe; A Sasaki, Several centers of atomic and molecular data in the world maintain research databases for use in fusion plasma simulations, hadron therapy, modelling the universe and other areas. Among the data center activities, collection of experimental and theoretical results across the world has been of major importance. This includes the identification, relevance assessment and retrieval of journal articles, followed by the data extraction, data mining, format conversion and data input. The methodology of the process still largely relies on working groups of specialists and part-time human labor, in spite of recent modernization in journal publishing, especially the electronic journals newly available in subscription domain and the free-access online abstract databases. This work focuses on automating the above procedure to the maximum extent possible. In particular, we design a download robot that performs query search and abstract retrieval for the candidates of relevant articles over the internet at first stage,. followed by fultext retrieval (pdf format), text extraction and a deterministic relevance judgement. As a demonstration, we have also developed a bibliography database for electron-molecule collisions that automatically updates its contents over the internet in regular time intervals. The present work belongs to the project for evolutional data collecting system supported by a JSPS project which involves several research institutes., 2005年, 3433, 159, 170, 研究論文(学術雑誌)
査読あり, 英語, PDPTA '05: Proceedings of the 2005 International Conference on Parallel and Distributed Processing Techniques and Applications, Vols 1-3, Extracting seismic electronic signals from the Telluric Current Data for the Nigata Chuetsu earthquake by ICA, C Ishikawa; C Watanabe; T Nagao; K Joe, This paper presents the results where we apply ICA(Independent Component Analysis) to TCD(Telluric Current Data.) observed in 2004 when the Nigata Chuetsu earthquake occurred, and detect SESs (Seismic Electric Signals). Detecting SESs from TCD is an effective method for short-term, earthquake prediction. In Japan, however, detecting SESs from TCD is considered as a difficult task because of train noise embedded in TCD. It becomes easy if we can separate train noise and SESs. Therefore, we investigate the application of ICA for the separation., 2005年, I, 488, 494, 研究論文(国際会議プロシーディングス)
査読あり, 英語, PDPTA '05: Proceedings of the 2005 International Conference on Parallel and Distributed Processing Techniques and Applications, Vols 1-3, Text classification for constructing an atomic and molecular journal database by LVQ, H Kashiwagi; C Watanabe; A Sasaki; K Joe, Atomic and molecular, data is essential in various research fields. The reliable data have been found in journal papers, and experts have tried to accumulate them as their own database. We can obtain necessary data out of online journals as a digital paper with the progress of web search engines. We need to classify the online papers to obtain the data. However, the classification is difficult because there are just abstracts available on web pages, or the number of papers to be download are limited. Using an LVQ algorithm, we propose a method to know whether a given online paper contains the atomic and molecular data using just the abstract. In this paper, we report a prototype system and its implementation as well as preliminary experimental results., 2005年, I, 481, 487, 研究論文(国際会議プロシーディングス)
査読あり, 英語, PDPTA '04: PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS 1-3, Visualization of seismic wave data by volume rendering and its application to an interactive query tool, A Uemura; C Watanabe; K Joe, We propose the Queryball model, which is a new style interaction model for a 3D visualization system. Queryball, which is a translucent ball, has search conditions, and two types of display methods. By locating a Queryball on the region of interest, users can observe just the objects they are interested in with emphasizing them and having other objects disappear. Users can change the region of interest by changing the area specification of the Queryball interactively. Although Queryball is useful for interactive analyze of various kinds of data, we need to consider and provide various usage of Queryball. In this paper, as the first step for providing such examples, we visualize seismic wave data and the underground structure of the earth. We apply some Queryballs to the results which are obtained by a volume rendering technique and an appropriate transfer function by examining the data distribution., 2004年, I, 366, 372, 研究論文(国際会議プロシーディングス)
査読あり, 英語, PDPTA '04: PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS 1-3, A query description model and its implementation as an interactive query tool for visualization systems, A Ishida; C Watanabe; K Joe, In this paper, we describe the development of a query definition support environment for Queryball, which is a new style of interaction tools developed at Nara Women's University. Queryball is a translucent ball, which has query conditions and visualization methods. Users can apply a query just inside the ball, and modify the area where the query is applied by moving and overlapping Queryballs. We propose several query models for Queryball using a description model named visualization pipeline. Based on the models, we developed a Queryball definition support environment with two description modes; one is to use a script language, the other is to use a visual programming tool. By using the visual programming tool, user can browse and edit the definitions of Queryball easily, intuitively, and graphically., 2004年, I, 359, 365, 研究論文(国際会議プロシーディングス)
査読あり, 日本語, 情報処理学会論文誌:数理モデル化と応用, iアプリを用いた数値計算の可能性, 城和貴; 高田雅美; 柴山智子; 渡辺知恵美; 庄野逸, 2004年, 46, SIG2, 47-55
査読あり, 英語, PDPTA '04: PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS 1-3, A multi-objective genetic algorithm for program partitioning and data distribution using TVRG, M Takata; T Yamaguchi; C Watanabe; Y Nakamura; K Joe, We propose an algorithm that performs data distribution and parallelization simultaneously. The objectives of the simultaneous algorithm are to reduce the length of critical path and the total memory size. Regardless to say, memory usage for each processor must be balanced. To obtain an optimal solution, we first adopted a branch and bound method. Since the branch and bound method often fails in the case of a large task graph, we adopt a multi-objective genetic algorithm, that provides a near optimal solution. For effective simultaneous partitionings, we employ some edge sorting and ordering methods. The effectiveness of our simultaneous partitioning algorithms is shown by experimental results., 2004年, I, 345, 351, 研究論文(国際会議プロシーディングス)
査読あり, 英語, 4th International Conference on Atomic and Molecular Data and Their Applications, Design and implementation of an evolutional data collecting system for the atomic and molecular databases, 城和貴, 2004年, 348-351
査読あり, 日本語, 情報処理学会論文誌:数理モデル化と応用, Webコミュニティ探索のための量子アルゴリズムの提案と検証, 城和貴, 2003年, 44, SIG7, 26-34
査読あり, 日本語, 情報処理学会論文誌:コンピュータシステム, 擬似スペクトル法を用いた乱流場の直接数値シミュレーションの並列化と性能評価, 城和貴, 2003年, 44, SIG6, 45-54
査読あり, 英語, PDPTA'03: PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS 1-4, Reduction and classification of input parameters for large-scale simulations, H Matsuyama; Y Matsuoka; M Koganeyama; C Watanabe; K Joe; Y Ueshima, Large-scale simulations in various fields are performed because they are available and inexpensive as a result of widespread use of high performance computers, networks and visualization methods. An integrated management system for simulation cycles has been developed. However, the existing management system often requires complex operation for data management. We investigate an autonomous agent system that advises the simulation users about management operation for large-scale simulations. Specifically, we develop an input parameter customizing agent which is a part of the agent system. As the first prototype of the agent, we analyze input parameter sets for large-scale simulations at JAERI Kansai Research Establishment Advanced Photon Research Center. In this paper, we propose an input parameter analysis method, input parameter sets are reduced to smaller dimensions using Principal Component Analysis (PCA), and classified using learning Vector Quantization (LVQ) according to simulation types. The validation of the input parameter analysis method is presented by experiments., 2003年, III, 1127, 1133, 研究論文(国際会議プロシーディングス)
査読あり, 英語, PDPTA'03: PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS 1-4, Queryball: A new model for querying in immersive VR systems, C Watanabe; A Osugi; Y Masunaga; K Joe, In this paper, we propose a powerful and convenient querying model, "Queryball", for users to query in immersive VR systems. In immersive VR systems, users expect intuitive and heuristic interactive operations derived from empirical knowledge in the real world. As a possible method for the interactive operations, interaction by virtual tools has been proposed. However, intuitive and heuristic query interaction by virtual tools is too difficult to be performed, because query interaction in immersive VR systems cannot be represented as concrete virtual objects.
Queryball is a new style of querying model for intuitive and heuristic query interaction. Queryball, which is a translucent ball, is defined as a quadruple of area specification, search condition, and two types of display methods. The search condition can be applied only to the virtual objects inside of the Queryball. The query results are shown to the user according to the display method. The position and radius of the Queryball changes when the user moves the Queryball interactively with observing the query results., 2003年, III, 1106, 1112, 研究論文(国際会議プロシーディングス)
査読あり, 日本語, 日本データベース学会 Letters, Queryball: 没入型VRシステムのための対話的な問合せモデル, 城和貴, 2003年, 2, 2, 25-28
査読あり, 英語, PDPTA'03: PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS 1-4, Specification of a parallel I/O control agent for large-scale simulation users, Y Matsuoka; H Matsuyama; M Koganeyania; C Watanabe; Y Ueshima; K Joe, An integrated management system for large-scale simulations saves the output data simulation results without considering the amount of empty space on the disks in the data servers. The present problem is inefficient operations on the data servers because each the disk sizes on the data server are unbalanced. We design and implement a prototype parallel I/O control agent for large-scale simulation data servers which supervises the data servers, makes suggestions to users and moves data flies between disks to keep disk sizes well-balanced., 2003年, III, 1120, 1126, 研究論文(国際会議プロシーディングス)
査読あり, 英語, 4th International Symposium on Independent Component Analysis and Blind Signal Separation, An Effective Evaluation Function for ICA to Separate Train Noise from Telluric Current Data, 城和貴, 2003年, 867-872
査読あり, 英語, The 2002 International Conference on Parallel and Distributed Processing Techniques and Applications, Design and Implementation of Unimodular Transformation on PROMIS, 城和貴, 2002年, III, 1438-1443
査読あり, 英語, The 2002 International Conference on Parallel and Distributed Processing Techniques and Applications, Quantum Algorithm for Searching Web Communities, 城和貴, 2002年, I, 260-266
査読あり, 日本語, 情報処理学会論文誌:数理モデル化と応用, 量子コンピュータ実現に向けた量子命令セットについて, 城和貴; 大音真由美; 中條拓伯; 高田司郎, 2002年, 43, SIG7, 19-28
査読あり, 日本語, 情報処理学会論文誌:数理モデル化と応用, ICAを用いた地電流データからの電車ノイズおよび地震前兆シグナルの分離, 城和貴; 小金山美賀; 庄野逸; 長尾年恭, 2002年, 43, SIG7, 92-104
査読あり, 英語, The 2002 International Conference on Parallel and Distributed Processing Techniques and Applications, An Improvement of Program Partitioning Based Genetic Algorithm, 城和貴, 2002年, I, 215-221
査読あり, 英語, The 2002 International Conference on Parallel and Distributed Processing Techniques and Applications,, Classification of Visualized Data Dependence, 城和貴, 2002年, III, 1444--1450
査読あり, 英語, The 2002 International Conference on Parallel and Distributed Processing Techniques and Applications, Parallelization of Seismic Wave Calculation by Impulse Response\nFunctions, 城和貴, 2002年, III, 1465-1471
査読あり, 英語, International Workshop on Advanced Compiler Technology for High Performance and Embedded Systems, Task and Variable Representation Graph: An Intermediate Representation of Parallelizing Compilers for Distributed Shared Memory Systems, 城和貴, 2001年, 47-55
査読あり, 英語, PDPTA'2001: PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, MIRAI: Automatic parallelizing and distributing compiler based on cc-COMA approach, T Uehara; T Nakanishi; M Mineo; S Saito; K Joe; A Fukuda; Y Kunieda, This paper describes the outline of the design and implementation of MIRAI, an automatic parallelizing and distributing compiler for PC clusters. The compiler is designed in cc-COMA framework, cc-COMA stands for "compiler-controlled cache only memory architecture" which is proposed as a basis for developing automatic parallelizing compilers. In this framework, the compiler generates object code along with the embedded software libraries to handle distributed shared memory provided by the runtime system. MIRAI generates object code for Fagus which is a software distributed shared memory system for a cluster of PCs. The compiler is also integrated with the visualization tool to allow users to restructure their program interactively., 2001年, 1193, 1199, 研究論文(国際会議プロシーディングス)
査読あり, 英語, PDPTA'2001: PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, A possible instruction set for quantum computer architectures, M Oto; H Nakajo; K Joe, The research of quantum computers, which is a noteworthy topic nowadays, consists of two major directions: 1) quantum devices/elements and 2) algorithms on quantum Turing machine. Although both are quite important and basic research for quantum computers, we believe another approach from computer architecture side is required to obtain real quantum computers. We have investigated the possible architecture for quantum computers. In this paper, we first propose a possible instruction set for quantum computer architectures. Using the instructions, we present programs of five well-known quantum algorithms. Based on the program model, we furthermore propose a processor architecture for quantum computers., 2001年, 1221, 1227, 研究論文(国際会議プロシーディングス)
査読あり, 英語, 5th International Conference and Exhibition on High-Performance computing in the Asia-Pacific Region, Effective Parallelization of a Turbulent Flow Simulation, 城和貴, 2001年, CD-ROM Proceedings
査読あり, 英語, PDPTA'2001: PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, Detecting seismic electric signals by LVQ based clustering, K Fukuda; M Koganeyama; H Shouno; K Joe; T Nagao, Aiming at short-term prediction of earthquakes, we have proposed the use of neural networks for analyzing telluric current data observed by the VAN method. We have already tried a telluric CUM-Cat data analysis method with learning Vector Quantization. In this paper, we will show preliminary experimental results for categorization of telluric current data by its frequency for the Izu islands earthquakes in Japan., 2001年, 1305, 1311, 研究論文(国際会議プロシーディングス)
査読あり, 英語, 3rd International Conference on Independent Component Analysis and Blind Signal Separation, Separation of Train Noise and Seismic Electric Signals from Telluric Current Data by ICA, 城和貴, 2001年, CD-ROM Proceedings
査読あり, 英語, PDPTA'2001: PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, The design and implementation of a UIR interface for the MIRAI parallelizing compiler, T Yamaguchi; H Shouno; K Joe, Since UIRs (Universal Intermediate Representations) for parallelizing compilers are a complicated arid the data set involved is quite large. Implementation, porting and maintenance of various optimizations are quite difficult, To overcome this problem, we propose a set of standard operations to a UIR of a parallelizing compiler as a UIR interface. In this paper, the specification and implementation issues of the UIR interface arc presented, and implementation examples of several tool, transformation methods are given and evaluated., 2001年, 1200, 1206, 研究論文(国際会議プロシーディングス)
査読あり, 英語, INNOVATIVE ARCHITECTURE FOR FUTURE GENERATION HIGH-PERFORMANCE PROCESSORS AND SYSTEMS, A heuristic approach to improve a branch and bound based program partitioning algorithm, M Takata; Y Kunieda; K Joe, In this paper we propose several heuristics that improve the branch and bound based program partitioning algorithm proposed by Girkar et. al., and evaluate the effectiveness by experiments. The heuristic depends heavily on the element order of edges of a given task graph. Therefore, it is necessary to sort the edges carefully to make effective use of the heuristic. Different sorting methods are investigated and experimentally evaluated. Approximate solutions that provide a sufficient practical partitioning were obtained using the accelerated heuristic, and execution times and error compared to the optimal solutions decreased considerably by sorting the edges of the task graph., 2000年, 105, 114, 研究論文(国際会議プロシーディングス)
査読あり, 英語, PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS I-V, Collaboration of Parafrase-2 and NaraView for effective parallelization supports, M Haneda; M Sasakura; U Nagashima; Y Kunieda; K Joe, To use parallel computer systems effectively, users need to reconstruct programs, and applications into parallel their sequential parallelization is not easy work for general users. Parallelizing compilers were developed to solve this problem but it is still quite difficult to parallelize applications correctly, efficiently, and automatically For this reason, some parallelization support tools are desired. NaraView, as one of the support tool, visualizes a given program by extracting internal information from a parallelizing compiler Parafrase-2. In this paper, we validate the usefulness of NaraView by parallelizing a real application., 2000年, 635, 641, 研究論文(国際会議プロシーディングス)
査読あり, 英語, PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS I-V, Removing train noise from telluric current data by neural networks for automatic short-term earthquake prediction in Japan, M Koganeyama; M Sugita; T Nagao; K Joe, The goal of this research is to obtain a practicable and reliable method for short-term earthquake prediction to detect seismic electric signals in telluric current data, which is used for the VAN method, by computers automatically. In this paper we perform experiments to know Cf neural networks can be applied to the reduction of train noise from telluric current data, which is considered as the main problem for the effective use of the VAN method in Japan and describe the experimental results., 2000年, 659, 665, 研究論文(国際会議プロシーディングス)
査読あり, 英語, PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS I-V, Accelerated program partitioning algorithm - An improvement of Girkar's algorithm, M Takata; Y Kunieda; K Joe, In this paper, we propose several heuristics that improve the branch and bound based program partitioning algorithm proposed by Girkar et. al., and indicate the effectiveness by experiments. The heuristic depends on the element order of edges of given task graphs deeply. Therefor, it is needed to sort edges pertinently for the effective use of the heuristic. The ways of sorting are investigated with experiments. As the results of the experiments, it turned out that approximate solutions, which provide enough practical partitionings, by accelerated heuristic is obtained, and execution times and errors to optimal solutions decrease considerably by sorting edges., 2000年, 699, 705, 研究論文(国際会議プロシーディングス)
査読あり, 英語, PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS I-V, Intuitive data partitioning by a simple physical model and its visualization, M Sugita; M Outo; M Sasakura; Y Kunieda; K Joe, For the efficient use of distributed memory systems, we face to data partitioning problem as well as program partitioning problem. That is how data and tasks are divided into distributed memory modules. In this paper, we propose intuitive data partitioning using the capability of human's visual recognition. To take full advantage of the capability, we use visualized information of given programs to be partitioned. In our visualization system, data and tasks are represented by an intermediate representation of a parallelizing compiler, and roughly clustered according to access costs described by a physical model between data and tasks. Users can decide a partitioning of data and tasks by looking at the visualized information. We show several example figures generated by our system., 2000年, 729, 735, 研究論文(国際会議プロシーディングス)
査読あり, 英語, The 4th International Conference on Supercomputing in Nuclear Applications, Automatic Short-term Earthquake Prediction by Neural Network - System Design and Expected Amount of Computation for Learning -, 城和貴, 2000年, CD-Rom Proceedings
査読あり, 英語, PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS I-V, Design and implementation of an automatic parallelizing and distributing compiler with visualization tools and the runtime environment, Y Kunieda; K Joe; A Fukuda; T Uehara; S Saito; T Saito; M Sasakura; T Nakanishi, This paper describes the outline of the research project to develop a fully-automatic parallelizing and distributing compiler for distributed-memory machines. The compiler is designed on. the framework of "Compiler-controlled cache only memory architecture (cc-COMA)" which is to develop the compiler to generate object code along with the embedded software libraries to handle distributed shared memory provided by the runtime system. To realize automatic data partitioning and distribution, Data Partitioning Graph (DPG) is introduced as the universal intermediate representation (UIR) in the compiler. The compiler is integrated with the visualization tool to allow interactive restructuring of the program by users. The runtime environment to realize software-controlled COMA runtime environment on networks of workstations (NOWs) is also provided., 2000年, 713, 719, 研究論文(国際会議プロシーディングス)
査読あり, 英語, PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS I-V, Wind: a Low-Costs Communication module for software DSM systems, S Saito; A Hayashi; T Uehara; K Joe; Y Kunieda, We are developing a Lour-Costs Communication module named "Wind" for software DSM(Distributed Shared Memory) systems. Wind is built in a Linux kernel, and can, copy directly each packet from the buffer for UDP (User Datagram Protocol) into the specified location, of virtual address space which assigned far a user process without any context switches between the kernel and the user process. Therefore Wind males it possible that user processes even, on a heavy load computer can receive every packet without loss. Wind is mainly aimed at transporting one page data of DSM. This paper shows an implementation and performance evaluations of Wind., 2000年, 721, 727, 研究論文(国際会議プロシーディングス)
査読あり, 英語, INNOVATIVE ARCHITECTURE FOR FUTURE GENERATION HIGH-PERFORMANCE PROCESSORS AND SYSTEMS, Laurasia and Wind: a compiler-controlled software DSM for WS/PC clusters, T Uehara; S Saito; K Joe; Y Kunieda, We have proposed an approach to build an automatic parallelizing compiler for various parallel computer architectures by providing COMA (Cache-Only Memory Architecture) environment that can be controlled by the compiler. This paper presents the design and implementation of a software DSM (Distributed Shared Memory) system named "Laurasia". It also describes a Low-Costs Communication module named "Wind" which is developed to accelerate the performance of Laurasia. Laurasia is designed to run on a cluster of UNIX Workstations or PCs, and Wind is built in a Linux kernel to eliminate context switches between the kernel and the user process., 2000年, 87, 93, 研究論文(国際会議プロシーディングス)
査読あり, 英語, INTERNATIONAL JOURNAL OF PARALLEL PROGRAMMING, NaraView: An interactive 3D visualization system for parallelization of programs, M Sasakura; K Joe; Y Kunieda; K Araki, For effective use of parallelizing compilers, an interactive environment which allows users to find more parallelism is needed. As the first step towards building such an environment, we have developed a program visualization system named NaraView. In this paper, we describe two visualization methods in NaraView. One is Program Structure View which illustrates the hierarchical loop structure of a given program and suggests which parts of the program can be parallelized. Another is the Data Dependence View which visualizes each data dependence on every variable or array element which is accessed in a specific loop. By using these views, users can easily understand which part of the program can be parallelized further. We also show several examples to demonstrate the efficiency of these methods., 1999年04月, 27, 2, 111, 129, 研究論文(学術雑誌)
査読あり, 英語, Proceedings of the Innovative Architecture for Future Generation High-Performance Processors and Systems, A heuristic approach to improve a branch and bound based - Program partitioning algorithm, Masami Takata; Yoshitoshi Kunieda; Kazuki Joe, In this paper, we propose several heuristics that improve the branch and hound based program partitioning algorithm proposed by Girkar et. al, and evaluate the effectiveness by experiments. The heuristic depends heavily on the element order of edges of a given task graph. Therefore, it is necessary to sort the edges carefully to make effective use of the heuristic. Different sorting methods are investigated and experimentally evaluated. Approximate solutions that provide a sufficient practical partitioning were obtained using the accelerated heuristic, and execution times and error compared to the optimal solutions decreased considerably by sorting the edges of the task graph., 1999年, 1999-, 105, 113, 研究論文(国際会議プロシーディングス), 10.1109/IWIA.1999.898848
査読あり, 日本語, 情報処理学会論文誌:数理モデル化と応用, リリース・コンシステンシ・モデルとその実現の形式的仕様記述について, 城和貴; 高田司郎; 田口研治; 福田 晃, 1999年, 40, SIG9, 1-17
査読あり, 英語, ACM ICS Workshop on Scheduling Algorithms for Parallel / Distributed Computing -- From Theory to Practice--, A Visualization Method for a Universal Intermediate Representation of Parallelizing Compilers, 城和貴, 1999年, 55-62
査読あり, 日本語, 情報処理学会論文誌:数理モデル化と応用, 変数オリエンティッドなデータ依存関係モデルの提案, 城和貴; 笹倉万里子; 國枝義敏; 荒木啓二郎, 1999年, 40, SIG2, 45-54
査読あり, 英語, Journal of Supercomputing, Object-oriented framework for loop parallelization, Youichi Omori; Akira Fukuda; Kazuki Joe, Generation of efficient parallel code is a major goal of a well-designed and developed parallelizing compiler. Another important goal is portability of both compiler system and the resulting output source codes. The various choices of current and future parallel computer architectures as well as the cost of developing a parallelizing compiler make portability a very important design goal. Since the design of parallelizing compilers is considerably move complex than designing conventional compilers, it is very important to achieve both efficiency and portability. To meet this dual goal, we have investigated the application of object oriented design to parallelizing compilers. Our parallelizing compiler design is based on abstractions of intermediate representations of loops and their class definitions. In this paper, we address the problem of loop parallelization and propose a framework where the loop parallelization process is divided into three phases and the optimization of loops is performed via a cyclic application of these three phases. The class of each phase is hierarchically derived from intermediate representations of loops. This facilitates the portability of the resulting parallelizing compilers. Furthermore, one of the phases uses a reservation table of hardware resources in order to obtain optimized parallel programs for given hardware resources. The validation of the proposed framework is given through the application of the object oriented design on an example program which is then parallelized efficiently., 1999年, 13, 1, 57, 69, 研究論文(学術雑誌), 10.1023/A:1008062717485
査読あり, 英語, International Conference on Parallel Architectures and Compilation Techniques, The Modulo Interval: A Simple and Practical Representation for Program Analysis, 城和貴, 1999年, 91-96
査読あり, 英語, ACM ICS Workshop on Scheduling Algorithms for Parallel / Distributed Computing -- From Theory to Practice--, Introducing Artificial Neural Networks to estimate lower bounds for a Branch and Bound Based Program Partitioning Algorithm, 城和貴, 1999年, 13-20
査読あり, 英語, 情報処理学会論文誌:数理モデル化と応用, A Combined Data and Program Partitioning Algorithm for Distributed Memory Parallel Computers, 城和貴, 1999年, 40, SIG9, 30-41
査読あり, 英語, INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS I-V, PROCEEDINGS, An implementation of loop transformations with a Universal Intermediate Representation interface library, K Kambe; T Nakanishi; K Joe; Y Kunieda; F Kako, This paper presents an implementation of a Universal Intermediate Representation (UIR) interface library for a parallelizing compiler Narafrase, and experimental evaluation results of the interface library through the implementation of two well-known loop transformations, loop distribution and loop permutation, are obtained. The UIR interface separates designers (who implement UIRs) and users (who implement program transformations). Therefore the interface library mal;es most program transformations independent of implementation, details of the UIR data structures, which encapsulate all information, necessary to exploit parallelism. Furthermore. we can obtain the portability of program transformations to other compiler environments with the interface library. By the experimental evaluation results, we prove that we can develop loop transformation techniques easily and efficiently with the interface library including not only a primitive library set (basic library functions) but also an. application, library set (sophisticated library functions)., 1999年, IV, 1905, 1911, 研究論文(国際会議プロシーディングス)
査読あり, 英語, INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS I-V, PROCEEDINGS, Converting different intermediate representations of parallelizing compilers: A case study, N Soyama; T Nakanishi; K Joe; Y Kunieda; F Kako, Many researchers of parallelizing compilers have proposed various design and implementation schemes of intermediate representations (IRs) for the last several decades. The foundation of ms seems to reach some common ideas, nevertheless they have actually spent much time to complete the foundation of ms individually. Recently the concept of universal IRs (UIRs) has been proposed for such general purpose schemes. According to the schemes, different IRs should be integrated even after Weir implementations, but it may be hard to obtain common UIRs practically. lit. this paper, we compare different Ins of two parallelizing compilers, then we show the possibility of the conversion of different Ins. As a result, we find that it is practically possible to convert different IRs except some implementation details of the IR data structures, and there are many similarities between different IRs., 1999年, IV, 1874, 1880, 研究論文(国際会議プロシーディングス)
査読あり, 英語, INNOVATIVE ARCHITECTURE FOR FUTURE GENERATION HIGH-PERFORMANCE PROCESSORS AND SYSTEMS, cc-COMA: the compiler-controled COMA as a framework for parallel computing, S Saito; T Uehara; K Joe; Y Kunieda, In order to provide a fully-automatic parallelzing compiler with the function of both automatic data partitioning and distribution, we propose a compiler-controlled cache only memory architecture (cc-COMA). The cc-COMA runtime system is based on a, software-emulated COMA which covers a variety of parallel architectures from NUMA to NOW. The compiler generates both the user code and control code for the emulated cache based on the Entry Consistency model. The runtime system of cc-COMA executes the generated code following the suggestions of the compiler. In this paper we present art outline of what we call cc-COMA system. It is also described about an implementation of the cc-COMA runtime system for networks of workstations, named Laurasia., 1999年, 114, 119, 研究論文(国際会議プロシーディングス)
査読あり, 英語, INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS I-V, PROCEEDINGS, The design and implementation of cc-COMA as a platform for distributed-memory machines, T Uehara; S Saito; K Joe; Y Kunieda, "Compiler-controlled cache only memory architecture (cc-COMA)" is proposed as a framework to develop fully-automatic parallelizing compilers for distributed-memory machines. A cc-COMA system consists of the runtime system to realize virtually shared memory like a COMA machine on the distributed memory and the compiler which generates object code along with the embedded primitives to maintain the cache coherency. This paper presents all experimental implementation of a cc-COMA runtime environment named "Laurasia". Laurasia realizes the software-controlled COMA runtime environment on networks of workstations (NOW's) along with the compiler interface to maintain coherency among the nodes. The experimental result shows that the performance of parallel programs can be dramatically improved by the embedded code inserted by the compiler that controls the cache coherency., 1999年, V, 1946, 1952, 研究論文(国際会議プロシーディングス)
査読あり, 英語, IEICE TRANSACTIONS ON INFORMATION AND SYSTEMS, Analytic modeling of updating based Cache coherent parallel computers, K Joe; A Fukuda, In this paper, we apply the Semi-markov Memory and Cache coherence Interference (SMCI) model, which we had proposed for invalidating based cache coherent parallel computers, to an updating based protocol. The model proposed here, the SMCI/Dragon model, can predict performance of cache coherent parallel computers with the Dragon protocol as well as the original SMCI model for the Synapse protocol. Conventional analytic models by stochastic processes to describe parallel computers have the problem of numerical explosion in the number of states necessary as the system size increases. We have already shown that the SMCI model achieved both the small number of states to describe parallel computers with the Synapse protocol and the inexpensive computation cost to predict their performance. In this paper, we demonstrate generality of the SMCI model by applying it to the another cache coherence protocol, Dragon, which has opposite characteristics than Synapse. We show the number of states required by constructing the SMCI/Dragon model is only 21 which is as small as SMCI/Synapse, and the computation cost is also the order of microseconds. Using the SMCI/Dragon model, we investigate several comparative experiments with widely known simulation results. We found that there is only a 5.4% differences between the simulation and the SMCI/Dragon model., 1998年06月, E81D, 6, 504, 512, 研究論文(学術雑誌)
査読あり, 英語, INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS I-IV, PROCEEDINGS, Specification and verification of memory consistency models for shared-memory multiprocessor systems, S Takata; K Taguchi; K Joe; A Fukuda, In this paper we will formally specify and verify memory consistency models for shared-memory multiprocessor systems, focusing on causal memory consistency model, by use of a formal specification technique proposed by Taguchi and Araki. The formal specification technique includes a language, which is based on the combination of the Z notation and CCS (Calculus of Communicating Systems), and the state-based CCS semantics, which integrates Z and CCS semantics. Causal memory requires that a read operation obtains the value that is consistent with other causally related read and write operations. A formal definition, implementation and verification of the causal memory have already been presented by Ahamad and Hutto. We will formally specify the behavior of the causal memory by the combination of Z and CCS and verify that the specified causal memory meets the causal memory consistency condition using the extended state-based CCS semantics., 1998年, 40, SIG2, 923, 930, 研究論文(国際会議プロシーディングス)
査読あり, 英語, INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS I-IV, PROCEEDINGS, NaraView: a visualization system for parallelization of programs, M Sasakura; K Joe; Y Kunieda; K Araki, For effective use of parallel computers, a tool which assists users to know the way of parallelization is needed. Since we believe visualization is a useful tool for parallelization, we are developing a tool named NaraView which visualizes the information from parallelizing compilers for helping users to understand programs. In this paper, we propose two visualization methods in NaraView, which are 3D visualization of program structure and data dependence., 1998年, 931, 932, 研究論文(国際会議プロシーディングス)
査読あり, 英語, INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS I-IV, PROCEEDINGS, Specification and verification of memory consistency models for shared-memory multiprocessor systems, S Takata; K Taguchi; K Joe; A Fukuda, In this paper we will formally specify and verify memory consistency models for shared-memory multiprocessor systems, focusing on causal memory consistency model, by use of a formal specification technique proposed by Taguchi and Araki. The formal specification technique includes a language, which is based on the combination of the Z notation and CCS (Calculus of Communicating Systems), and the state-based CCS semantics, which integrates Z and CCS semantics. Causal memory requires that a read operation obtains the value that is consistent with other causally related read and write operations. A formal definition, implementation and verification of the causal memory have already been presented by Ahamad and Hutto. We will formally specify the behavior of the causal memory by the combination of Z and CCS and verify that the specified causal memory meets the causal memory consistency condition using the extended state-based CCS semantics., 1998年, 2, 923, 930, 研究論文(国際会議プロシーディングス)
査読あり, 英語, INTERNATIONAL CONFERENCE ON PARALLEL AND DISTRIBUTED PROCESSING TECHNIQUES AND APPLICATIONS, VOLS I-IV, PROCEEDINGS, NaraView: a visualization system for parallelization of programs, M Sasakura; K Joe; Y Kunieda; K Araki, For effective use of parallel computers, a tool which assists users to know the way of parallelization is needed. Since we believe visualization is a useful tool for parallelization, we are developing a tool named NaraView which visualizes the information from parallelizing compilers for helping users to understand programs. In this paper, we propose two visualization methods in NaraView, which are 3D visualization of program structure and data dependence., 1998年, 27, 2, 931, 932, 研究論文(国際会議プロシーディングス)
査読あり, 英語, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Naraview: An interactive 3d visualization system for parallelization of programs, Mariko Sasakura; Kazuki Joe; Keijiro Araki, For effective use of parallelizing compilers, an interactive environment which allows users to instruct the way of parallelization is needed. As the first step to build such an environment, we have developped a program visualization system named Nara View. The system provides two powerful methods for 3D visualization of program structure and data dependence. 3D visualization of program structure illustrates a hierarchical loop structure of given programs and suggests which parts of the program have been parallelized. 3D visualization of data dependence explains each data dependence on any variable or array element which is accessed at a specific loop. By using these methods, users can easily understand which part of the program should be more parallelized. We also show several examples to demonstrate the efficiency of these methods., 1997年, 1336, 231, 242, 研究論文(国際会議プロシーディングス), 10.1007/BFb0024219
査読あり, 英語, COMPSAC 97 : TWENTY-FIRST ANNUAL INTERNATIONAL COMPUTER SOFTWARE & APPLICATIONS CONFERENCE, A parallelizing compiler by object oriented design, Y Omori; K Joe; A Fukuda, Applying conventional compiler design methodology to a parallelizing compiler, ids internal data structures tend to get too complicated quickly, Thus, we introduce object oriented design from the problem analysis stage and achieve the following improvements: 1) consistent modeling from the theory to the implementation, 2) reduced program sire through improved reusability based on better class design methodology, and 3) flexible coding through a stub class far parallelization. We extract objects based on stream and thus clarify similarities and differences to the conventional design schemes. Then, we show the framework of the internal classes used in our parallelizing compiler which enhances design-level reusability in C++ implementation of the compiler. Furthermore, toe provide a virtual class to be used as a unit of MIMD style parallel execution and mate it a common representation among different parallelization algorithms. Finally, we compare our design against SUIF, which abo uses C++, and show the improvements on the design classes., 1997年, 232, 239, 研究論文(国際会議プロシーディングス)
査読あり, 英語, Journal of Supercomputing, Object-oriented framework for loop parallelization, Youichi Omori; Akira Fukuda; Kazuki Joe, Generation of efficient parallel code is a major goal of a well-designed and developed parallelizing compiler. Another important goal is portability of both compiler system and the resulting output source codes. The various choices of current and future parallel computer architectures as well as the cost of developing a parallelizing compiler make portability a very important design goal. Since the design of parallelizing compilers is considerably move complex than designing conventional compilers, it is very important to achieve both efficiency and portability. To meet this dual goal, we have investigated the application of object oriented design to parallelizing compilers. Our parallelizing compiler design is based on abstractions of intermediate representations of loops and their class definitions. In this paper, we address the problem of loop parallelization and propose a framework where the loop parallelization process is divided into three phases and the optimization of loops is performed via a cyclic application of these three phases. The class of each phase is hierarchically derived from intermediate representations of loops. This facilitates the portability of the resulting parallelizing compilers. Furthermore, one of the phases uses a reservation table of hardware resources in order to obtain optimized parallel programs for given hardware resources. The validation of the proposed framework is given through the application of the object oriented design on an example program which is then parallelized efficiently., 1999年, 13, 1, 57, 69, 研究論文(学術雑誌), 10.1023/A:1008062717485
査読あり, 英語, IEICE TRANSACTIONS ON INFORMATION AND SYSTEMS, Analytic modeling of cache coherence based parallel computers, K Joe; A Fukuda, In this paper, we propose an analytic model using a semi-markov process for parallel computers which provides hardware support for a cache coherence mechanism. The model proposed here, the Semi-markov Memory and Cache coherence Interference model, can be used for the performance prediction of cache coherence based parallel computers since it can be easily applied to descriptions of the waiting states due to network contention or memory interference of both normal data accesses and cache coherence requests. Conventional analytic models using stochastic processes to describe parallel computers have the problem of numerical explosion in the number of states necessary as the system size increases even for simple parallel computers without cache coherence mechanisms. The number of states required by constructing our proposing analytic model, however, does not depend on the system size but only on the kind of cache coherence protocol. For example, the number of states for the Synapse cache coherence protocol is only 20, as is described in this paper. Using the proposed analytic model, we investigate several comparative experiments with widely known simulation results. We found that there is only a 7.08 % difference between the simulation and our analytic model, while our analytic model can predict the performance of a 1,024 processor system in the order of microseconds., 1996年07月, E79D, 7, 925, 935, 研究論文(学術雑誌)
査読あり, 英語, The 1996 International Conference on Parallel and Distributed Processing Techniques and Applications, A Data Block Mapping Method to Reduce Cache Coherence Overhead, 城和貴, 1996年, 2, 691-701
査読あり, 英語, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Estimating minimum execution time of perfect loop nests with loop-carried dependences, Tsuneo Nakanishi; Kazuki Joe; Constantine D. Polychronopoulos; Keijiro Araki; Akira Fukuda, We propose a scheme to estimate exact minimum parallel execution time of perfect loop nests with loop-carried dependences at iteration and instruction-level parallelism. We formulate the problem of the estimation as an integer programming problem and solve it with a branch-and-bound scheme combined with the simplex method. Execution time obtained with the proposed scheme is useful to evaluate effects of applications of various optimization or parallelizing techniques for iteration or instruction-level oarallel execution of loons., 1997年, 1239, 97, 599, 研究論文(国際会議プロシーディングス), 10.1007/BFb0017281
査読あり, 英語, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Estimating minimum execution time of perfect loop nests with loop-carried dependences, Tsuneo Nakanishi; Kazuki Joe; Constantine D. Polychronopoulos; Keijiro Araki; Akira Fukuda, We propose a scheme to estimate exact minimum parallel execution time of perfect loop nests with loop-carried dependences at iteration and instruction-level parallelism. We formulate the problem of the estimation as an integer programming problem and solve it with a branch-and-bound scheme combined with the simplex method. Execution time obtained with the proposed scheme is useful to evaluate effects of applications of various optimization or parallelizing techniques for iteration or instruction-level oarallel execution of loons., 1997年, 1239, 97, 599, 研究論文(国際会議プロシーディングス), 10.1007/BFb0017281
査読あり, 英語, Lecture Notes in Computer Science 1124, Applying the Semi-Markov Memory and Cache Coherence Interference Model to an Updating Based Cache Coherence Protocol, 城和貴, 1996年, 706-713
査読あり, 英語, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 3D visualization of program structure and data dependence for parallelizing compilers and parallel programming, Mariko Sasakura; Satoko Kiwada; Kazuki Joe; Tsuneo Nakanishi; Keijiro Araki, For effective use of parallel computers, a tool which assists users to know the way of parallelization is needed. Since we believe visualization is a useful tool for parallelization, we are developing a tool named NaraView for parallelizing Fortran programs. In this paper, we propose two visualization methods in NaraView, which are 3D visualization of program structure and data dependence., 1997年, 1239, 591, 593, 研究論文(国際会議プロシーディングス), 10.1007/BFb0017279
査読あり, 英語, International Symposium on Parallel and Distributed Supercomputing, An Analysis of the SCIDDLE Library on a Workstation Cluster: A Step Towards a Globally Distributed Locally Parallel Processing Environment, 城和貴, 1995年, 104-111
査読あり, 英語, International Conference on Parallel Processing, CDPP Algorithm: Combined Data and Program Partitioning, 城和貴, 1995年, II, 177-181
査読あり, 英語, International Conference on Parallel Architectures and Compilation Techniques, A Loop Parallelization Technique for Linear Dependence Vector, 城和貴, 1995年, 285-289
査読あり, 英語, High Performance Computing Conference'94, The Semi-Markov Memory and Cache Coherence Interference Model, 城和貴, 1994年, 254-265
査読あり, 英語, International Computer Symposium, A Parametric Simulator For a Parallel Computer Using Address-Trace, 城和貴, 1994年, 979-985
査読あり, 英語, Lecture Notes of Computer Science 892, The Data Partitioning Graph: Extending Data and Control Dependencies for Data Partitioning, 城和貴, 1994年, 170-185
査読あり, 英語, 2nd International Workshop on Massive Parallelism: Hardware, Software and Applications, An Analytic Model for a Hierarchical Parallel System, 城和貴, 1994年, 287-304
査読あり, 英語, SUPERCOMPUTING '93, PROCEEDINGS, A DISTRIBUTED SHARED-MEMORY MULTIPROCESSOR - ASURA - MEMORY AND CACHE ARCHITECTURES, S MORI; H SAITO; M GOSHIMA; M YANAGIHARA; T TANAKA; D FRASER; K JOE; H NITTA; S TOMITA, 1993年, 740, 749, 研究論文(国際会議プロシーディングス)
査読あり, 英語, International Workshop on Performance Evaluation on Parallel Systems, An Analytic Model for a Parallel Computer, -- Prediction of Shared Block Behavior --, 城和貴, 1993年, 67-74
査読あり, 英語, IJCNN INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS, VOLS 1-3, APOLONN BRINGS US TO THE REAL WORLD - LEARNING NONLINEAR DYNAMICS AND FLUCTUATIONS IN NATURE, M SATO; K JOE; T HIRAHARA, 1990年, A581, A587, 研究論文(国際会議プロシーディングス)
査読あり, 英語, 2nd Neural Information Processing System, A Large-Scale Neural Network Which Recognizes Hanwritten Kanji Characters, 城和貴, 1990年, 415-422
査読あり, 英語, Concurrency: Practice and Experience, Construction of a large‐scale neural network: Simulation of handwritten Japanese character recognition on NCUBE, Kazuki Joe; Yoshihiro Mori; Sei Miyake, This paper describes how new learning methods may make it possible for a large‐scale, hierarchical neural network to recognize most Japanese handwritten characters. This is a very large and complex task, as the Japanese character set consists of about 3000 categories which can be written in many different ways. Such a difficult task can lead a neural network to converge very slowly and to yield recognition rates that are uneven between categories. To address these problems we here propose five learning methods as modifications of the conventional back‐propagation learning rule. These methods produce fast convergence, even recognition rates over all categories, and adequate recognition of test samples. We also describe how a large‐scale neural network can be built by dividing the recognition task into several subtasks, with networks for each subtask, and then integrating these subnetworks in a large network with a hierarchical structure. In a hierarchical network, the upper level network directly integrates outputs from each lower level network. Application of that network to handwritten Japanese character recognition has resulted in poor recognition, because lower level networks do not know about unknown input patterns, and the direct integration of ambiguous outputs from many lower level networks confuses the upper level network. We propose a new integration method which provides each subnetwork with more information as to how close an input pattern is to the categories of that subnetwork. This method resulted in high recognition performance for character recognition. We here described the above methods, and report the performance of our implementation of a neural network for the recognition of 71 Hiragana characters, and describe our implementation of this network on a hypercube concurrent computer. Copyright © 1990 John Wiley &
Sons, Ltd, 1990年, 2, 2, 79, 107, 研究論文(学術雑誌), 10.1002/cpe.4330020202
査読あり, 英語, International Conference on Fuzzy Logic and Neural Networks, Learning Chaotic Dynamics by Recurrent Neural Networks, 城和貴, 1990年, 601-604
査読あり, 英語, 4th Conference of Hypercube Concurrent Computers and Applications, Simulation of a Large-Scale Neural Network on a Parallel Computer -- An Application for Handwritten Japanese Character Recognition--, 城和貴, 1989年, 2, 1111-1118
英語, 生物物理, 一般社団法人 日本生物物理学会, 3P046 動きの特徴を用いたタンパク質構造の時系列解析(蛋白質-構造機能相関,第48回日本生物物理学会年会), Kamada Mayumi; Toda Mikito; Sekijima Masakazu; Takata Masami; Joe Kazuki, 2010年, 50, 2, S153, 10.2142/biophys.50.S153_2
査読あり, 英語, Proceedings of the Innovative Architecture for Future Generation High-Performance Processors and Systems, Laurasia and wind : A compiler-controlled software DSM for WS/PC clusters, Tetsutaro Uehara; Shoichi Saito; Kazuki Joe; Yoshitoshi Kunieda, We have proposed an approach to build an automatic parallelizing compiler for various parallel computer architectures by providing COMA (Cache-Only Memory Architecture) environment that can be controlled by the compiler. This paper presents the design and implementation of a software DSM (Distributed Shared Memory) system named "Laurasian. It. also describes a Low-Costs Communication module named "Wind" which is developed to acccleratc the performance of Laurasia. Laumsia is designed to run on a cluster of UNIX Workstations or PCs, and Wind is built in a Linux kernel to eliminate context switches between the kernel and the user process., 1999年, 1999-, 87, 93, 研究論文(国際会議プロシーディングス), 10.1109/IWIA.1999.898846
査読あり, その他, The 2021 International Conference on Parallel and Distributed Processing Techniques and Applications, Estimating Concrete Interior using 3D-CNN, Miho Chiyonobu; Sayaka Iida; Yu Ishikawa; Masaharu Nishikino; Ryuta Kawakami; Kazuki Joe, 2023年, 研究論文(国際会議プロシーディングス)
査読あり, その他, 情報処理学会論文誌数理モデル化と応用, 多段組多サイズ見出しで構成される近代書籍のレイアウト解析, 飯田紗也香; 竹本有紀; 石川由羽; 髙田雅美; 城和貴, 2023年, 研究論文(学術雑誌)
査読あり, その他, The 2022 International Conference on Parallel and Distributed Processing Techniques and Applications, Application of Deep Metric Learning to Early-modern Japanese Printed Character Recognition, Norie Koiso; Yuki Takemoto; Sayaka Iida; Yu Ishikawa; Masami Takata; Kazuki Joe, 2022年, 研究論文(国際会議プロシーディングス)
査読あり, その他, The 2022 International Conference on Parallel and Distributed Processing Techniques and Applications, Translating Early-modern Written Style into Current Colloquial Style in Hoji Shinbun, Honoka Nishikawa; Yuki Takemoto; Sayaka Iida; Yu Ishikawa; Masami Takata; Kaoru Ueda; Kazuki Joe, 2022年, 研究論文(国際会議プロシーディングス)
査読あり, その他, 情報処理学会論文誌数理モデル化と応用, 特定の近代書籍出版者における低出現頻度文字種の獲得方法, 竹本有紀; 石川由羽; 高田雅美; 城和貴, 2022年, 15, 3, 71, 89, 研究論文(学術雑誌)
査読あり, その他, The 2020 International Conference on Parallel and Distributed Processing Techniques and Applications, Crawling Low Appearance Frequency Characters Images for Early-Modern Japanese Printed Character Recognition, Nanami Fujisaki; Yu Ishikawa; Masami Takata; Kazuki Joe, 2021年, 683, 695, 研究論文(国際会議プロシーディングス)
日本語, 情報処理学会数理モデル化と問題解決研究会, 情報処理学会, 速度低下錯視を用いた VR 酔いの低減手法, 2020年-2月-2日, 127, 11, 1, 6
日本語, 情報処理学会数理モデル化と問題解決研究会, 情報処理学会, GPを用いたJAZZ実時間自動生成システムの試作, 2020年-1月2-日, 131, 18, 1, 6
日本語, 情報処理学会数理モデル化と問題解決研究会, 情報処理学会, 邦字新聞における近代文語体と現代口語体の自動翻訳の検討, 2020年-1月2-日, 131, 12, 1, 6
日本語, 情報処理学会数理モデル化と問題解決研究会, 情報処理学会, レーザー打音分析によるコンクリート内部状態の推定, 2021年-2月-2日, 132, 14, 1, 6
日本語, 情報処理学会数理モデル化と問題解決研究会, 情報処理学会, 近代書籍における文字切り出し手法の検討, 2021年-2月-2日, 132, 4, 1, 6
査読無し, 英語, 5TH BIOMEDICAL ENGINEERING INTERNATIONAL CONFERENCE (BMEICON 2012), IEEE, Constitution and Phase Analysis of Alpha Waves, Yu Ishikawa; Masami Takata; Kazuki Joe, 2012年, 98, 98, 研究発表ペーパー・要旨(国際会議)
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 一般社団法人情報処理学会, SVMを用いた生体分子への金属結合部位予測手法の提案, 中澤 昌美; 高田 雅美; 横田 恭宣; 野口 保; 関嶋 政和; 城 和貴, タンパク質は,生体内に存在する金属イオンと結合することで立体構造が変化し,それに伴い生体内での機能が起こる.金属タンパク質はさまざまな生理機能の発現に深く関与し,生命活動において特徴のある非常に重要な役割を果たしている.現在タンパク質の立体構造はX線結晶構造解析やNMRによって解析が進められているが,これらの実験手法は金属の結合情報を得るまでに相当な時間を要する.そこで本研究では,既知の金属イオンとタンパク質の結合情報をタンパク質の立体構造データベースであるPDB(Protein Data Bank)から抽出し,SVMによる機械学習を用いて,金属イオンの結合性を予測する手法と結合部位を予測する手法を提案する., 2009年02月26日, 73, 19, 213, 216
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 一般社団法人情報処理学会, タンパク質立体構造における時系列特徴抽出, 木村 紗知; 戸田 幹人; 関嶋 政和; 高田 雅美; 野口 保; 城 和貴, 熱揺らぎをしているタンパク質は時々刻々と周波数が変化している可能性がある.ウェーブレット変換を用いて,周波数変化を解析することで熱揺らぎを受けたタンパク質の動きを捉えることができる.本研究では,経時的に変化するタンパク質を揺らぎのパターンでクラスタリングすることを目指す., 2009年02月26日, 73, 19, 97, 100
査読無し, その他, 第16 回「ハイパフォーマン スコンピューティングとアーキテクチャの評価」に関する北海道ワークショップ(HOKKE-2009)2009-ARC-174/HPC-119, Cell/B.E.とDIMMnetを併用 した大容量ボリュームレンダリングの並列処理性能, 田邊昇; 佐々木愛美; 中條拓伯; 高田雅美; 城和貴, 2009年, -
査読無し, 日本語, 電子情報通信学会技術研究報告. CPSY, コンピュータシステム, 社団法人電子情報通信学会, 大容量データ向け対話的実時間遠隔可視化装置の実現性検討(ストレージと可視化,SWoPP佐賀2008-2008年並列/分散/協調処理に関する『佐賀』サマー・ワークショップ), 田邊昇; 佐々木愛美; 中條拓伯; 城和貴, スーパーコンピュータが生成する大容量データの可視化を、対話的に遠隔から行えるようにすることが必要とされている。我々は数十GBのボリュームデータを保持し、これを実時間で可視化すると同時に、映像をH.264のライブストリーミングで配信し、遠隔から対話的に制御できるシステムを提案する。本報告では、上記のシステムを2年後に構築することを仮定し、その実現性を検討する。単一のGPUに装備されているメモリ容量を大幅に超えるボリュームデータのサイズでは、GPUから見てローカルではなくホストインタフェースの向こう側のデータをアクセスする確率が高くなる。ストリーミング用のエンコーディングも必要である。そこで、通常のGPUの並列利用の可能性のみならず、東芝のSpurs Engineの並列利用も検討する。ボリュームレンダリングにおける視線方向のメモリアクセスはキャッシュフレンドリーではない。ボリュームレンダリングを行なう際の大容量メモリへのアクセスのアーキテクチャ的サポートについても検討する。また,ユーザの視点の位置に伴い、必ずしも常に最高解像度が必要とされない。レスポンスとの兼ね合いから、データの間引きについても検討する。, 2008年07月29日, 108, 180, 43, 48
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, E-Cell System を用いたアルコール耐性酵母に関するモデル構築, 鎌田 真由美; 仲嶋 なつ; 石川 千里; 高田 雅美; 城 和貴, 2008年07月14日, 70, 27, 30
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 自由視点映像における背景モデル作成のための一手法, 一色彩; 坂本竜基; 北原格; 小暮潔; 城和貴, 自由視点映像における背景領域の形状モデルを簡便に生成する手法を提案する.前景領域の自由視点映像は,視体積交差法などの手法により3次元形状を復元することで実現されている.一方で,背景領域の形状モデルはCADなどの3Dモデリングソフトを用いて作成されることが多い。この場合,詳細なモデルの作成には手間がかかり,また,簡易的に作成した単純な形状のモデルを使用すると映像が不自然なものとなっていた.本提案手法では,背景領域の形状を複数の平面で近似することとし,多視点画像に写り込んだ平面の頂点を対応点として指定するという簡単な処理により背景モデルを生成する.各対応点の3次元座標は,前景領域の3次元推定に用いられる多視点カメラのキャリブレーション情報を基にステレオ処理により算出される,実際の形状に比較的近いモデルを簡易に生成することが可能である., 2008年07月07日, 2008, 65, 1, 4
査読無し, 英語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, アナグリフ動画ライブストリーミング配信の実現, 小堀由貴; 石川千里; 高田雅美; 城和貴, 近年,パソコンの革新的な発展に伴い,立体画像は世間の脚光を浴びている.高性能計算機によって,画像処理はもはや難しい処理ではない.このような状況において,立体動画像は特に注目を集めている.最近のパソコンはとても高性能であるためリアルタイム動画生成は十分行えるようになっている.さらに,近年ブロードバンドネットワークは様々なコンテンツのストリーミングを可能としている.特に,動画のライブストリーミングは様々な分野で活用されている.本稿では,リアルタイムでアナグリフによる3D動画像を生成し,ライブストリーミング配信するシステムを提案する.提案システムでは,2台のUSBカメラを使用することにより,それぞれのキャプチャ画像が,リアルタイムで立体画像に合成するためにフィルタリングされ,生成された3D動画像が即座にライブストリーミング配信される.また実験によって,自ら提案したシステムが十分実用的であることも示す., 2008年07月07日, 2008, 65, 5, 8
査読無し, 英語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, E-Cell Systemを用いたアルコール耐性酵母に関するモデル構築, 鎌田真由美; 仲嶋なつ; 石川千里; 高田雅美; 城和貴, 清酒醸造では特有の醸造手法を用いるため酵母菌の生成するエタノールは高濃度に達する.そのため醸造工程においてアルコールに強い耐性を持つものが求められる.このような酵母菌は時間,コストのかかる試行錯誤的な実験によって取得される.そこで既知の文献などを元にアルコール耐性酵母菌に関して計算機上でシミュレーションモデルの構築を試み,そのモデルを用いて新たな知見獲得を試みる.本稿では,汎用細胞シミュレーションシステムであるE-Cell Systemを用いて酵素菌の細胞壁の主成分である1,3-βグルカンの簡潔な合成代謝経路,シグナル伝達としてPkc1-MAPKカスケードに関してモデルの構築を試みる., 2008年07月07日, 2008, 65, 27, 30
査読無し, 日本語, 電子情報通信学会技術研究報告. WIT, 福祉情報工学, 社団法人電子情報通信学会, 書字運動の訓練支援ソフトウェアの開発(福祉と音声処理,一般), 芳野可奈子; 高田雅美; 天白成一; 城和貴, 我々は,書字運動に困難を抱える児童の書字訓練用に新たにソフトウェアを開発している.従来から利用されている典型的な訓練方法に,紙の上で文字をなぞり書きをする手法があるが,個々の能力に適した対応をするには難しい.新たに開発している書字学習用のリハビリテーション用のソフトウェアは,携帯型ゲーム機ニンテンドーDS上で動作するため,多彩な学習プログラムの自動提供,ならびに紙上で行うよりも反復学習が容易である.さらに,ゲーム性を加えることで学習者の興味を持続できることも利点である.本稿では,書字運動の訓練手法の提案とそれを踏まえた開発中のソフトウェアについて述べる., 2008年05月22日, 108, 67, 79, 83
査読無し, その他, 電子情報通信学会技術研究報告. SP, 音声, 社団法人電子情報通信学会, 書字運動の訓練支援ソフトウェアの開発(福祉と音声処理,一般), 芳野可奈子; 高田雅美; 天白成一; 城和貴, 我々は,書字運動に困難を抱える児童の書字訓練用に新たにソフトウェアを開発している.従来から利用されている典型的な訓練方法に,紙の上で文字をなぞり書きをする手法があるが,個々の能力に適した対応をするには難しい.新たに開発している書字学習用のリハビリテーション用のソフトウェアは,携帯型ゲーム機ニンテンドーDS上で動作するため,多彩な学習プログラムの自動提供,ならびに紙上で行うよりも反復学習が容易である.さらに,ゲーム性を加えることで学習者の興味を持続できることも利点である.本稿では,書字運動の訓練手法の提案とそれを踏まえた開発中のソフトウェアについて述べる., 2008年05月22日, 108, 66, 79, 83
査読無し, その他, 情報処理学会論文誌. 数理モデル化と応用, 社団法人情報処理学会, 「情報処理学会論文誌:数理モデル化と応用」の編集にあたって, 城和貴, 2008年03月15日, 49, 4
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, E-Cell System を用いた酵母菌のストレス応答MAPK経路のモデル構築, 仲嶋 なつ; 鎌田 真由美; 石川 千里; 高田 雅美; 城 和貴, 2008年03月04日, 68, 1, 4
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, リアルタイム Queryball の開発に向けて, 勅使河原; 佑美; 来見 真理; 石川 千里; 高田 雅美; 城 和貴, 2008年03月04日, 68, 5, 8
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, E-Cell Systemを用いた酵母菌のストレス応答MAPK経路のモデル構築, 仲嶋なつ; 鎌田真由美; 石川千里; 高田雅美; 城和貴, 日本酒の製造過程において高濃度アルコールを生成するためには,発酵を行う酵母菌にアルコール耐性を持つものが適するとされている.そのような酵母菌を発見するために様々な実験が行われているが,実際の培養には酵母菌死滅を含む多くの代償が伴う.そこでこのような代償を少なくするため,酵母菌のアルコール耐性に関するシミュレーションモデルを構築し,アルコール耐性に関する新たな知見の獲得を試みる.本研究ではアルコール耐性と酵母菌細胞壁との関係に着目し,E-Cell Systemを用いた細胞壁合成Pkc1-MAPK経路の部分的なモデル構築を行う., 2008年03月04日, 2008, 17, 1, 4
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, リアルタイムQueryballの開発に向けて, 勅使河原佑美; 来見真理; 石川千里; 高田雅美; 城和貴, 3次元空間内のオブジェクトを理解するための機能の1つとしてQueryballがある.Queryballはオブジェクトに半透明の球体を重ね合わせることにより,オブジェクトと球体が重なる部分に対して問合せることができ,オブジェクトの表示方法を変えたり,様々な方向からオブジェクトを視認したりできる.しかし,現段階のQueryballのボリュームレンダリングでの表示速度はオブジェクトデータのサイズが大きくなるほど遅くなり,実用には向かない.そこで,Queryballをリアルタイムで実行するために,ボリュームレンダリング部分にGPGPUを適用することによって,高速化を図る., 2008年03月04日, 2008, 17, 5, 8
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, GPを用いた指向性フィルタの生成と評価, 美坂千穂; 石川千里; 高田雅美; 城和貴, 近年のデジタルデバイスの進歩に伴い,画像処理技術の活躍分野が広がっている.現在用いられている画像処理のアルゴリズムは処理対象が限定され,汎用性が無いものが多い.そこで,GPを用いて汎用性のある画像処理フィルタを作成する研究に着手した.本稿では,画像処理フィルタの中でも基本的なエッジ検出フィルタに着目し,GPによる自動生成を試みる,先行実験として,基本的なGPの適用実験を行った.得られたフィルタを適用した結果は一見良好であったが,エッジの方向によって検出できないものも存在した.よって,本稿ではGPを利用して適切なエッジの指向性を学習させる手法を提案する., 2008年03月04日, 2008, 17, 21, 24
査読無し, その他, The first International Workshop on Super Visualization (IWSV’08) in conjunction with the 22nd ACM International Conference on Supercomputing (2008), Feasibility Study of a Remote Interactive Real-Time Visualization System Using Live Streaming, Noboru Tanabe; Hironori Nakajo; Kazuki Joe, 2008年, -
査読無し, その他, 情報処理学会研究報告. BIO, バイオ情報学, 社団法人情報処理学会, 特定ユーザのための嗜好分析パターン抽出の一手法, 児玉理映子; 石川千里; 高田雅美; 城和貴, ユーザの嗜好に合致する商品を提示するレコメンド法が注目されている.しかしながら,従来のレコメンド法では全てのユーザを満足させることはできない.そこで,満足できないと考えられる特定ユーザの嗜好を抽出する新たなレコメンド法を提案する.本稿では,特定ユーザとして興味の有無が明確なオタクを採用した.オタクの嗜好を抽出するにあたってまずオタクの定義をし,その定義に基づいて選出したユーザをオタクとする.またオタク以外のユーザを一般人とする.このオタクと一般人のWebアクセスログから決定木を生成しネツト上での行動パターンを抽出し分析する.分析結果からオタクは一般人と比べ膨大な情報量を持つコンテンツを閲覧していることが示された., 2007年12月20日, 2007, 128, 61, 64
査読無し, 日本語, 情報処理学会研究報告. BIO, バイオ情報学, 社団法人情報処理学会, 複数Webページの重要文抽出および直感的理解を支援するためのGUIの開発, 柴田裕子; 山内和子; 石川千里; 高田雅美; 城和貴, 近年,Web空間は情報収集における重要な情報源のひとつとなった.しかし,Web空間には多種多様な情報が氾濫しているため,検索エンジンを用いて必要な情報を得ようとしても検索結果の数は膨大であり,ユーザが情報過多による混乱を起こす恐れがある.そこで本研究では,より効率的に必要な情報を取得するため,検索結果から得られる複数Web文書から重要文とキーワードを抽出するモデルを提案する.本稿では,モデルの提案と同時にユーザの直感的理解を視覚的に支援するGUIの開発を行った.本モデルでは,検索結果として得られる複数の文書に目を通さなければならないという人間にかかる負担を軽減させることを目的としている., 2007年12月20日, 2007, 128, 81, 84
査読無し, 日本語, 情報処理学会研究報告. BIO, バイオ情報学, 社団法人情報処理学会, 3Dスキャナのための自動テクスチャマッピング手法, 持永彩華; 樽井沙知; 石川千里; 高田雅美; 城和貴, 3Dスキャナを使用した3Dオブジェクト化の作業には多くのコストを要する.そこで本稿では,3Dオブジェクトに対して,テクスチャマッピングを施す位置を自動検出する手法を提案する.テクスチャマッピングを施す位置選択において,テクスチャとして用いる写真内の画像形状が,3Dオブジェクトと一致する位置へのマッピングがより最適であると考えられる.本手法では,3Dオブジェクトのレンダリング画像と,実物の写真画像とのマスク情報をマッチングすることで,テクスチャマッピングを行う位置を自動的に検出する.マッチング計算にSSDA法を適用することにより,計算の高速化を図る.評価実験として,貝標本の3Dオブジェクトに自動マッピングを施す., 2007年12月20日, 2007, 128, 85, 88
査読無し, その他, 情報処理学会研究報告. BIO, バイオ情報学, 社団法人情報処理学会, リアルタイムアナグリフ動画像生成, 小堀由貴; 白川有莉; 石川千里; 高田雅美; 城和貴, 近年,映画やゲーム,アートから遊園地のアトラクションに至るまでの実にさまざまな分野で三次元可視化技術が利用されている.本稿では,そのような可視化技術のなかでも特に立体視に注目し,その実現のために2台のUSBカメラから撮影した映像をOpenCVを用いてリアルタイムでアナグリフ映像に変換して出力することを試みる.またOpenCVによる処理速度低下の欠点を克服するために,映像(動画)や音声を扱うためのフレームワークであるDirectShowをキャプチャ部分にのみ取り入れて,変換処理の高速化を図る., 2007年12月20日, 2007, 128, 89, 92
査読無し, 日本語, 情報処理学会研究報告. BIO, バイオ情報学, 社団法人情報処理学会, E-cell systemを用いた酵母菌のグルカン合成に関するモデル構築, 鎌田真由美; 仲嶋なつ; 石川千里; 高田雅美; 城和貴, 清酒醸造工程では,特有の製法で高濃度アルコールが生成される.清酒の発酵を行う酵母菌には,アルコール耐性を持つものが適しており,そのような酵母菌を発見するために現在様々な研究や実験が行われている.その中で,酵母菌のアルコール耐性については,遺伝子レベルで徐々に明らかになってきている.しかしながら,実際の酵母菌を用いた実験には,多くのコストを必要とする.そこで,アルコール耐性に関するシミュレーションモデルを構築し,アルコール耐性に関する新たな知見の獲得を試みる.本稿では,まず細胞壁の主成分であるグルカンに着目し,汎用細胞シミュレーションソフトであるE-Cell Systemを用いて,1,3-βグルカン合成に関する簡潔な代謝経路についてモデル構築を行う., 2007年12月20日, 2007, 128, 113, 120
査読無し, 日本語, 情報処理学会研究報告. BIO, バイオ情報学, 社団法人情報処理学会, 地球流体科学研究用のポータルGfdnaviへの3次元可視化機能の組み込み, 来見真理; 当麻英梨子; 石川千里; 高田雅美; 城和貴, 近年,幅広い分野で可視化技術は必要不可欠となっており,特に科学技術の分野ではシミュレーションや計測の結果をわかりやすく表示するために2次元図や3次元図が用いられている.本稿では,地球流体科学研究用のポータルとして開発中であるGfdnaviに3次元可視化機能を組み込みX-windowのリモート機能を利用して結果をクライアント側に出力する.さらに,3次元可視化機能をGPGPUで高速化を図ることにより,より使いやすいソフトウェアの開発を考える., 2007年12月20日, 2007, 128, 163, 166
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 特定ユーザのための嗜好分析パターン抽出の一手法, 児玉理映子; 石川千里; 高田雅美; 城和貴, ユーザの嗜好に合致する商品を提示するレコメンド法が注目されている.しかしながら,従来のレコメンド法では全てのユーザを満足させることはできない.そこで,満足できないと考えられる特定ユーザの嗜好を抽出する新たなレコメンド法を提案する.本稿では,特定ユーザとして興味の有無が明確なオタクを採用した.オタクの嗜好を抽出するにあたってまずオタクの定義をし,その定義に基づいて選出したユーザをオタクとする.またオタク以外のユーザを一般人とする.このオタクと一般人のWebアクセスログから決定木を生成しネツト上での行動パターンを抽出し分析する.分析結果からオタクは一般人と比べ膨大な情報量を持つコンテンツを閲覧していることが示された., 2007年12月20日, 2007, 128, 61, 64
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 複数Webページの重要文抽出および直感的理解を支援するためのGUIの開発, 柴田裕子; 山内和子; 石川千里; 高田雅美; 城和貴, 近年,Web空間は情報収集における重要な情報源のひとつとなった.しかし,Web空間には多種多様な情報が氾濫しているため,検索エンジンを用いて必要な情報を得ようとしても検索結果の数は膨大であり,ユーザが情報過多による混乱を起こす恐れがある.そこで本研究では,より効率的に必要な情報を取得するため,検索結果から得られる複数Web文書から重要文とキーワードを抽出するモデルを提案する.本稿では,モデルの提案と同時にユーザの直感的理解を視覚的に支援するGUIの開発を行った.本モデルでは,検索結果として得られる複数の文書に目を通さなければならないという人間にかかる負担を軽減させることを目的としている., 2007年12月20日, 2007, 128, 81, 84
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 3Dスキャナのための自動テクスチャマッピング手法, 持永彩華; 樽井沙知; 石川千里; 高田雅美; 城和貴, 3Dスキャナを使用した3Dオブジェクト化の作業には多くのコストを要する.そこで本稿では,3Dオブジェクトに対して,テクスチャマッピングを施す位置を自動検出する手法を提案する.テクスチャマッピングを施す位置選択において,テクスチャとして用いる写真内の画像形状が,3Dオブジェクトと一致する位置へのマッピングがより最適であると考えられる.本手法では,3Dオブジェクトのレンダリング画像と,実物の写真画像とのマスク情報をマッチングすることで,テクスチャマッピングを行う位置を自動的に検出する.マッチング計算にSSDA法を適用することにより,計算の高速化を図る.評価実験として,貝標本の3Dオブジェクトに自動マッピングを施す., 2007年12月20日, 2007, 128, 85, 88
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, リアルタイムアナグリフ動画像生成, 小堀由貴; 白川有莉; 石川千里; 高田雅美; 城和貴, 近年,映画やゲーム,アートから遊園地のアトラクションに至るまでの実にさまざまな分野で三次元可視化技術が利用されている.本稿では,そのような可視化技術のなかでも特に立体視に注目し,その実現のために2台のUSBカメラから撮影した映像をOpenCVを用いてリアルタイムでアナグリフ映像に変換して出力することを試みる.またOpenCVによる処理速度低下の欠点を克服するために,映像(動画)や音声を扱うためのフレームワークであるDirectShowをキャプチャ部分にのみ取り入れて,変換処理の高速化を図る., 2007年12月20日, 2007, 128, 89, 92
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, E-cell systemを用いた酵母菌のグルカン合成に関するモデル構築, 鎌田真由美; 仲嶋なつ; 石川千里; 高田雅美; 城和貴, 清酒醸造工程では,特有の製法で高濃度アルコールが生成される.清酒の発酵を行う酵母菌には,アルコール耐性を持つものが適しており,そのような酵母菌を発見するために現在様々な研究や実験が行われている.その中で,酵母菌のアルコール耐性については,遺伝子レベルで徐々に明らかになってきている.しかしながら,実際の酵母菌を用いた実験には,多くのコストを必要とする.そこで,アルコール耐性に関するシミュレーションモデルを構築し,アルコール耐性に関する新たな知見の獲得を試みる.本稿では,まず細胞壁の主成分であるグルカンに着目し,汎用細胞シミュレーションソフトであるE-Cell Systemを用いて,1,3-βグルカン合成に関する簡潔な代謝経路についてモデル構築を行う., 2007年12月20日, 2007, 128, 113, 120
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 地球流体科学研究用のポータルGfdnaviへの3次元可視化機能の組み込み, 来見真理; 当麻英梨子; 石川千里; 高田雅美; 城和貴, 近年,幅広い分野で可視化技術は必要不可欠となっており,特に科学技術の分野ではシミュレーションや計測の結果をわかりやすく表示するために2次元図や3次元図が用いられている.本稿では,地球流体科学研究用のポータルとして開発中であるGfdnaviに3次元可視化機能を組み込みX-windowのリモート機能を利用して結果をクライアント側に出力する.さらに,3次元可視化機能をGPGPUで高速化を図ることにより,より使いやすいソフトウェアの開発を考える., 2007年12月20日, 2007, 128, 163, 166
査読無し, その他, 情報処理学会論文誌. 数理モデル化と応用, 社団法人情報処理学会, 情報爆発時代に向けた大気科学分野のための3次元可視化モデルの提案, 当麻英梨子; 高田雅美; 城和貴, 本稿では,Shneidermanが提唱する情報可視化に求められる7種の主要な操作機能を基にした3次元可視化モデルを提案する.提案する可視化モデルでは,情報爆発が生じている分野の1つである大気科学分野を対象とし,大量のデータ分析を可能とする.そのために,全体像と詳細情報の2種類の表示を用いる.この2種類の表示を効果的に使い分けることによって,大量のデータの中から効率良く特徴的現象を発見することが可能となる.インタフェースにおいては,Shneidermanが提唱する対話設計における8つの黄金律を採用する.提案モデルの有効性を示すために,モデルの適用によって開発した大気科学分野のための3次元可視化システムを用い,実際に生じた大気現象の可視化を行う., 2007年12月15日, 48, 19, 86, 98
査読無し, その他, 情報処理学会論文誌. 数理モデル化と応用, 社団法人情報処理学会, アブストラクトを用いた原子分子物理学分野の論文分類支援システムの設計と実装, 柏木裕恵; 高田雅美; 佐々木明; 城和貴, 原子・分子と電子の衝突による電離・励起の断面積データは様々な分野で活用され,原子分子物理学分野の論文に記載されている.ゆえに,これらのデータを利用するためには原子分子物理学分野の数多くの論文から情報を収集し,目的のデータが掲載されている論文を探し出す必要がある.論文は一般にオンラインジャーナルより入手するが,論文の閲覧が有料である等の問題から原子分子物理学分野のすべての論文を入手することは実質不可能である.そこで本論文では無料で入手でき,誰でも閲覧可能であるアブストラクトを用いて,データが掲載されている論文を探し出す支援をするシステムを開発する.従来,論文分類手法として多く用いられている機械学習法を本システムにも適用し,アブストラクトだけを用いて論文を分類することが可能であることを検証する., 2007年10月15日, 48, 15, 189, 198
査読無し, 日本語, 電子情報通信学会ソサイエティ大会講演論文集, 社団法人電子情報通信学会, AS-2-4 ICA適用による地電流データからの地震前駆的シグナル抽出結果に対する考察(AS-2.電磁波地震予知における信号処理の新展開、評価技術,シンポジウム), 石川千里; 高田雅美; 長尾年恭; 城和貴, 2007年08月29日, 2007, "S, 22"-"S-23"
査読無し, 日本語, 電子情報通信学会ソサイエティ大会講演論文集, 社団法人電子情報通信学会, AS-2-5 地震短期予測のための地電流解析Webシステムの構築(AS-2.電磁波地震予知における信号処理の新展開、評価技術,シンポジウム), 豊島良美; 石川千里; 高田雅美; 長尾年恭; 城和貴, 2007年08月29日, 2007, "S, 24"-"S-25"
査読無し, 英語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, SVMを用いた論文分類システムの構築, 芦田尚美; 高田雅美; 佐々木明; 鴨浩靖; 新出尚之; 城和貴, 本論文は,SVMを用いた論文分類システムの構築を目的とする.Web上には膨大な数の論文が存在し,研究者は自分が必要とするごく少数の論文をその中から探し出す必要がある.また,通常Webから無償で入手することが出来るのはアブストラクト部分のみであるので,必要な論文であるか否かを本文を読まずに判断しなければならない.そこで,本論文では論文のアブストラクト部分のみを用いて論文を自動分類するシステムの構築を試みる.本論文では,分類器としてSVMを用いた論文分類システムを構築し,先行研究との比較を行い,提案手法の有効性を示す., 2007年06月25日, 2007, 64, 21, 24
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 健康状態の推定を目的とした排便物識別のための一手法, 小松原美穂; 石川千里; 高田雅美; 鴨浩靖; 新出尚之; 城和貴, 本論文において,便の形状を分類するシステムを提案する.このシステムには,チェインコードと楕円フーリエ記述子を用いる.再現率と誤認識率を調べることによって,提案システムの評価を行う.実験の結果,再現率は75%以上となった., 2007年06月25日, 2007, 64, 25, 28
査読無し, 英語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 大気科学のための3次元可視化システムGateau : デザインコンセプトと実用評価, 当麻英梨子; 野口克行; 高田雅美; 鴨浩靖; 新出尚之; 城和貴, 3次元可視化を独自で作成するためには専門的知識を必要とするため,大気科学者はデータ分析において2次元可視化を利用している.そこで,我々は可視化プログラミングに精通していない大気科学者を対象とした3次元可視化システムGateauの開発を行っている.Gateauは大気科学分野で扱われている様々なデータフォーマットを読み込み可能とし,3次元可視化を用いて従来のデータ分析の過程を再現し,目的の現象の発見へと導く.さらに,その可視化結果はプレゼンテーションなどに直接利用可能である.本稿では,大気科学分野における従来の可視化手法と提案する3次元可視化システムの概要を説明し,実用評価を行う., 2007年06月25日, 2007, 64, 29, 32
査読無し, 英語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, Cellプロセッサへの分子動力学シミュレーションの最適化, 佐々木愛美; 関嶋政和; 高田雅美; 城和貴, 生体分子の機能解析に利用される分子動力学シミュレーションは,疾病に関与するタンパク質の揺らぎの原理を明らかにし,そのタンパク質と結合するペプチドを発見することを期待されている.このようなシミュレーションは,典型的な科学シミュレーションのように大規模で複雑な計算を必要とする.さらに,分子動力学シミュレーションは,フェムト秒単位での正確な計算に基づいている.本論文では,計算時間を短縮するために,分子動力学シミュレーションをPLAYSTATION3のCell Broadband Engine上で実行し,報告する., 2007年06月25日, 2007, 64, 37, 40
査読無し, その他, 情報処理学会論文誌. 数理モデル化と応用, 社団法人情報処理学会, 「情報処理学会論文誌 : 数理モデル化と応用」の編集にあたって, 城和貴, 2007年03月15日, 48, 6
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 食事と健康状態の関連を知るための一手法(Session 4), 李丹陽; 高田雅美; 城和貴, 近年,科学進歩に伴い,記憶装置の大容量化が進んでいる.その結果として,蓄積された大量のデータの中から得られる情報は,多種多様かつ複雑である.そのため,従来の統計解析手法では扱うことが難しいデータや,様々な形式のデータベースから,有用な情報を取り出す必要がある.このための技術として,データマイニングが注目されている.本論文では,摂取した食品と健康状態のデータに対してデータマイニングを適用することで,摂取した食品と健康状態の関連について人間の先入観を介入させず発見し,健康状態の管理に役立つ指標を作るために,食事と健康状態の関連を調べる手順を提案する.また,データマイニングを用いて食事と健康状態の相関ルールを発見するための実験を行う., 2007年03月03日, 2007, 19, 77, 80
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, ニンテンドーDSを用いた書字学習トレーニングソフトの開発(Session 4), 芳野可奈子; 高田雅美; 天白成一; 城和貴, 我々は,学習障害(LD)の中でも書字に問題があるDyslexiaの訓練用に新たにソフトウェアを開発している. Dyslexiaの障害を持つ人は,文字の読み書きの障害であり,従来から利用されている典型的な訓練方法は紙の上で文字をなぞり書きをすることのみであった.新たに開発している書字学習用のリハビリテーション用のソフトウェアは任天堂のDSという携帯型のゲーム機上で動作する.ゲーム機上で行うため紙上で行うよりも反復学習が容易である.さらに,ゲーム性を加えることで学習者の興味を持続できることも利点である.本稿では,開発背景と現在開発中のソフトウェアの内容について述べる., 2007年03月03日, 2007, 19, 81, 84
査読無し, その他, 情報処理学会研究報告, 情報処理学会, Auto View Selectionを実現する一手法, 中川真智子; 高田雅美; 城和貴, 2006年12月21日, 2006, 135, 49, 52
査読無し, その他, 情報処理学会研究報告. BIO, バイオ情報学, 社団法人情報処理学会, Auto View Selectionを実現する一手法(セッション3), 中川真智子; 高田雅美; 城和貴, 近年科学技術及び可視化技術の発達に伴い,多様な三次元可視化表示が行えるようになった.膨大で複雑なデータの可視化が可能になったことで,有益な知見を得るための視点の選択は重要となってくる.しかし,視点選択の定義は目的や対象によって異なってくるため,規格化された選択基準は存在しない,本稿では既存のモデルを改良し,複数の要素を組み合わせる事で視点の自動選択を行う., 2006年12月21日, 2006, 135, 49, 52
査読無し, その他, 情報処理学会研究報告. BIO, バイオ情報学, 社団法人情報処理学会, 短期地震研究のための統合的なマルチデータベースの設計(セッション3), 豊島良美; 石川千里; 高田雅美; 城和貴, 本稿では,既存の地電流を用いたオンラインシステムに,データベース機能を追加するための設計について述べる.地震による被害を減少させるためには,高精度な短期地震予測が不可欠である.予測を高精度に行うためには,様々な観測データを使用しなければならない.この際,いかに膨大なデータを効率的に扱えるかが問題となるが,多くのデータを効率的に扱うためにはデータベースを作成して利用すれば良い., 2006年12月21日, 2006, 135, 65, 68
査読無し, 日本語, 情報処理学会研究報告. BIO, バイオ情報学, 社団法人情報処理学会, 熱揺らぎを考慮したドッキングシステムの開発(セッション5), 木村紗知; 佐々木愛美; 関嶋政和; 高田雅美; 野口保; 城和貴, ドッキングシミュレーションは,タンパク質が別のタンパク質やペプチドなどと結合する現象をシミュレーションするものである.しかし,既存のドッキングプログラムはいずれも生体内で揺らいでいるタンパク質を氷のように固まったものとして扱うことが多く,結合するリガンドの挙動を正しく評価することが難しい.本稿では,この問題を回避し,生体内での揺らいでいるタンパク質とリガンドの結合を求めるために,分子動力学シミュレーションとドッキングシミュレーションを組み合わせたシステムの開発手法についての提案をおこなう., 2006年12月21日, 2006, 135, 119, 126
査読無し, その他, 情報処理学会研究報告. BIO, バイオ情報学, 社団法人情報処理学会, SVMを用いた金属イオン結合部位予測システムの開発(セッション6), 中澤昌美; 高田雅美; 横田恭宣; 野口保; 関嶋政和; 城和貴, 金属イオンの結合は,タンパク質の構造や機能の変化に大きな影響を与える.ゲノム配列がシーケンサーにより網羅的に読解されていくのに対して,タンパク質の立体構造はNMRやX線結晶解析によって解かれることから,あるゲノム配列が得られてから立体構造や金属イオン結合情報が明らかになるまでには相当の時間が要される.そこで本研究では,既知の金属イオンとタンパク質の結合情報をデータベースから抽出し,SVMによる機械学習から,タンパク質と金属イオンとの結合予測を行うシステム開発をしている., 2006年12月21日, 2006, 135, 157, 164
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, Auto View Selectionを実現する一手法(セッション3), 中川真智子; 高田雅美; 城和貴, 近年科学技術及び可視化技術の発達に伴い,多様な三次元可視化表示が行えるようになった.膨大で複雑なデータの可視化が可能になったことで,有益な知見を得るための視点の選択は重要となってくる.しかし,視点選択の定義は目的や対象によって異なってくるため,規格化された選択基準は存在しない,本稿では既存のモデルを改良し,複数の要素を組み合わせる事で視点の自動選択を行う., 2006年12月21日, 2006, 135, 49, 52
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 短期地震研究のための統合的なマルチデータベースの設計(セッション3), 豊島良美; 石川千里; 高田雅美; 城和貴, 本稿では,既存の地電流を用いたオンラインシステムに,データベース機能を追加するための設計について述べる.地震による被害を減少させるためには,高精度な短期地震予測が不可欠である.予測を高精度に行うためには,様々な観測データを使用しなければならない.この際,いかに膨大なデータを効率的に扱えるかが問題となるが,多くのデータを効率的に扱うためにはデータベースを作成して利用すれば良い., 2006年12月21日, 2006, 135, 65, 68
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 熱揺らぎを考慮したドッキングシステムの開発(セッション5), 木村紗知; 佐々木愛美; 関嶋政和; 高田雅美; 野口保; 城和貴, ドッキングシミュレーションは,タンパク質が別のタンパク質やペプチドなどと結合する現象をシミュレーションするものである.しかし,既存のドッキングプログラムはいずれも生体内で揺らいでいるタンパク質を氷のように固まったものとして扱うことが多く,結合するリガンドの挙動を正しく評価することが難しい.本稿では,この問題を回避し,生体内での揺らいでいるタンパク質とリガンドの結合を求めるために,分子動力学シミュレーションとドッキングシミュレーションを組み合わせたシステムの開発手法についての提案をおこなう., 2006年12月21日, 2006, 135, 119, 126
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, SVMを用いた金属イオン結合部位予測システムの開発(セッション6), 中澤昌美; 高田雅美; 横田恭宣; 野口保; 関嶋政和; 城和貴, 金属イオンの結合は,タンパク質の構造や機能の変化に大きな影響を与える.ゲノム配列がシーケンサーにより網羅的に読解されていくのに対して,タンパク質の立体構造はNMRやX線結晶解析によって解かれることから,あるゲノム配列が得られてから立体構造や金属イオン結合情報が明らかになるまでには相当の時間が要される.そこで本研究では,既知の金属イオンとタンパク質の結合情報をデータベースから抽出し,SVMによる機械学習から,タンパク質と金属イオンとの結合予測を行うシステム開発をしている., 2006年12月21日, 2006, 135, 157, 164
査読無し, その他, 情報処理学会論文誌. 数理モデル化と応用, 社団法人情報処理学会, 可視化されたデータ依存の分類とその並列化手法への適用, 山口智美; 笹倉万里子; 城和貴, 自動並列化コンパイラの研究では,長い間,ループ変換理論やデータ依存解析の理論的な研究が行われてきた.一方,並列システムは共有メモリ,分散メモリ,クラスタ,Gridと様々なトレンドが出現しており,自動並列化コンパイラがすべてに実装されるのを期待することは現実的ではない.現に多くの研究者が各々のアプリケーションの並列最適化を行うのに,MPI, OpenMP, HPF等の環境を使っている.ところが並列プログラミング初学者には,ループ変換理論やデータ依存解析を直接利用するには敷居が高く,自動並列化コンパイラ研究で培われてきた研究成果を利用できないという問題点があった.本論文では,並列プログラミング初学者がデータ依存関係を目視により直観的に理解し,得られたパターンからループ変換手法を対話的に通用する手法について提案し,その具体例について報告する., 2006年10月15日, 47, 14, 179, 186
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, アブストラクトを用いた論文分類システムの設計と実装(セッション3), 柏木裕恵; 高田雅美; 佐々木明; 城和貴, 本論文では,アブストラクトを用いた論文分類システムを提案する.オンラインで必要な論文を検索する際,アブストラクトの情報のみ閲覧可能である場合が多く,必要な論文を見つけ出すことは非常に困難である.従来の論文分類システムでは分類に論文が必須であるため,論文を入手していない論文検索段階では分類を行うことができない.しかし,我々の提案するシステムでは,アブストラクトさえあれば論文の分類が可能となるため,論文を検索する際に分類を行い,必要な論文だけを入手することができる.我々は,分類手法としてLVQを採用し,これまで困難とされてきた原子分子物理学分野の論文の分類実験を試みる.実験結果より,アブストラクトのみで論文分類を行うことができる提案手法の有効性を示す., 2006年09月14日, 2006, 95, 33, 36
査読無し, 英語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 地震短期予測のための統合的データベースシステムの設計, 石川千里; 鴨浩靖; 新出尚之; 城和貴, 本稿は,地震短期予測のための統合的データベースシステムの設計について述べるものである.現在,様々な電磁気学的手法を用いた地震短期予測が考えられている.これらの結果を統合的に調査し,地震と電磁気学的な値のデータとの相関関係を発見するためには,統合的なデータベースが必要である.これまで,我々は地電流に対してICAを適用することで地震短期予測を実現するという研究を行ってきたため,本稿では,統合的データベースの設計について,地電流に対してICAを適用する部分を中心に述べる., 2006年06月26日, 2006, 68, 27, 30
査読無し, 英語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, Gateau:大気科学研究のための直感的な分析を実現する3次元可視化システム, 当麻英梨子; 山内和子; 鴨浩靖; 新出尚之; 林田佐智子; 城和貴, 近年ハードウェアの進展に伴って,VolumeRenderingのような3D可視化技術が様々な分野で取り入れられている.大気科学分野でもCGが使われているが,主流は2次元可視化である.そこで我々は,大気科学研究のための3次元可視化システムGateauの開発を行っている.Gateauは可視化プログラミングに精通していない大気科学研究者を対象とし,3次元格子データを少ない手間で3次元可視化を実現するクイックルックツールである.マニュアルを必要とせず,シンプルなインターフェースを基調にインタラクティブに使い方をマスターすることができ,衛星観測データの標準配布フォーマットとなりつつなるHDF-EOSを読み込み可能とする., 2006年06月26日, 2006, 68, 31, 34
査読無し, 英語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, HDF-EOSのためのGateau APIの設計と実装, 山内和子; 当麻英梨子; 鴨浩靖; 新出尚之; 林田佐智子; 城和貴, 我々が開発しているGateauは,大気科学研究に用いられている衛星観測データを3次元可視化することで大気科学研究者がデータを分析することを支援するためのシステムである.本研究ではGateauに読み込ませる衛星観測データフォーマットとして,近年配布が増加しているHDF-EOSを選択した.大気科学研究者はHDF-EOS読み込み機能を使用することで,HDF-EOSライブラリと呼ばれる独自のAPIを用いたプログラムを記述することなく3次元可視化を行うことができる.本稿ではGateauへのHDF-EOS読み込み機能の組み込みについて述べる., 2006年06月26日, 2006, 68, 35, 38
査読無し, 英語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 構造化文書の3次元可視化と効果的なナビゲーションによる癒し系RSSリーダーの開発, 原聡美; 鴨浩靖; 新出尚之; 城和貴, 近年のblogの広がりに伴ってRSSやRSSリーダーが認知され利用も広がりつつある.現在多く用いられているRSSリーダーは,シンプルで事務的な外観・形式のものが多い.そこで,本研究では「楽しさ」や「癒し」をコンセプトとした新しいタイプのRSSリーダーの開発を行った.RSSの持つ階層構造に着目し,そうした構造を柱とその周りをとりまく球体とで3次元空間に可視化する.そうした空間を利用し,2つの新たな機能を持ったRSSリーダーの作成を行った.本稿ではそうしたRSSリーダーの機能として,ユーザが操作することがなくパソコンの余剰時間を利用できるアプリケーションとして,周囲の人間へ受動的視聴を促す「RSSViewer」という機能と,能動的な選択手段にさらに「癒し」「楽しさ」についても考慮し,空間内を視点が移動することによって記事を選択することができる「RSSWalker」という機能について述べる., 2006年06月26日, 2006, 68, 39, 42
査読無し, 日本語, 日本データベース学会letters, 日本データベース学会, 網羅的なシミュレーションにおける空間的パターン発見のためのデータマイニング手法の開発, 石田愛; 渡辺知恵美; 城和貴, 2006年06月, 5, 1, 145, 148
査読無し, 日本語, 言語処理学会年次大会発表論文集, 科学論文からの原子分子データの収集,データベース化の自動化支援に関する研究, 佐々木明; 村田真樹; 柏木裕恵; 城和貴; ピフル ルカーシュ; 加藤太治; 村上泉, 2006年03月13日, 12th, 943, 946
査読無し, その他, 情報処理学会論文誌. 数理モデル化と応用, 社団法人情報処理学会, 「情報処理学会論文誌: 数理モデル化と応用」の編集にあたって, 城和貴, 2005年12月15日, 46, 17
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 光量子科学研究におけるLVQを用いたデータ分類・可視化処理の自動化(セッション1), 上島豊; 斎藤寛二; 松山仁美; 城和貴, 光量子科学研究において, 研究過程で生じる画像データを分類したり, 多次元データを適切に可視化したりすることは研究者にとって大きな負担である.にもかかわらず, 研究者が管理しなければならないデータは増え続けるばかりである.一方, データの分類や可視化のノウハウは, 研究者個人のみに蓄積され続けている.それゆえ, データの管理, 分類, 可視化, 保全性の品質は, 全く均一ではなくなっている.その結果, 過去のデータ処理を再現させようとしても, その保証はどこにもないのが現状である.我々は, このような問題を解消するために, ニューラルネットワークの一種であるLearning Vector Quantization (LVQ)を利用して研究者のノウハウを知識化し, 画像の分類や適切な解析を研究者が行い易くするための画像データ解析支援システムの開発と評価を行った., 2005年09月21日, 2005, 93, 1, 4
査読無し, 英語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, スペクトル特性に基づいたQuery-by-Exampleによる音楽検索モデル(セッション1), YUYi; 渡辺知恵美; 城和貴, 近年, 内容ベースの音楽検索における研究はますます多くの関心を引き付けている.適切な特徴セットを用いた類似検索アプローチにより, 計算時間を減少させ検索速度を向上させることができる.本論文は音響ベースの音楽検索に対し以下の二点で貢献している: 1.スペクトル特性を研究し, 隣接しているフレームのスペクトル特徴が非常に関連していることを示している;2.基本的な音響特性の分析に基づき, スペクトル相関関係に焦点を合わせた高速で効率的なQuery-by-Exampleによる音楽検索モデルを提案している.また本提案手法の評価としてシミュレーション結果におけるスペクトル相関関係(SC)閾値, データ格納量, 計算時間の分析を行い, 単旋律および多声音楽における有効性を確認した., 2005年09月21日, 2005, 93, 5, 8
査読無し, 日本語, 原子衝突研究協会研究会講演概要集, 計算機による原子分子物理学の研究情報のデータベース構築の研究, 佐々木明; 上島豊; 村田真樹; 柏木裕恵; 城和貴; PICHL Lukas; 村上泉; 加藤太治; 加藤隆子, 2005年08月23日, 30th, 81, 82
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, ICAを用いた新潟中越地震前駆的シグナルの抽出(災害), 石川千里; 長尾年恭; 城和貴, 本稿は、新潟県中越大地震が発生した2004年1年間に観測された地電流データに、信号分離の技術であるICAを適用し、地震前駆的シグナルの検出を試みた結果について報告するものである。地震の短期予知は、地震前駆的シグナルと呼ばれる矩形波の電流を検出することで可能になるといわれているが、我が国では地電流からの矩形波検出は、同じく地電流に含まれる電車のノイズの影響により困難とされている。しかし、地電流の電車ノイズ成分と矩形波との分離が可能となれば、地震前駆的シグナルの検出も容易になるはずである。我々は地電流にICAを適用することで、その地電流の成分分離を試みている。, 2005年03月09日, 2005, 20, 57, 61
査読無し, その他, 情報処理学会論文誌. 数理モデル化と応用, 社団法人情報処理学会, iアプリを用いた数値計算の可能性, 高田雅美; 柴山智子; 渡辺知恵美; 庄野逸; 城和貴, 携帯電話会社NTT Docomoの提供するiアプリは,携帯端末において,様々なJavaアプリケーションを実行することが可能であるiアプリ対応の携帯端末は,計算機としては貧弱ではあるが,ネットワーク機能.計算リソースを有しており,2003年11月の時点では,その市場稼動台数は,2,000万台を突破している.本研究では,アイドル状態にある携帯端末において,iアプリを用いた数値計算を行うための分散処理システムによって,計算リソースを有効活用する方法を提案する.個々め携帯端末は,貧弱な計算リソースしか持たないため,処理すべき計算問題は,分散化される必要がある.そこで,我々は,この分散処理のための1つのアプローチとして分散遺伝的アルゴリズムをiアプリによって実装し,計算問題解決のためのプロトタイプを作成した., 2005年01月15日, 46, 2, 47, 55
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, Query Painter : 対話的可視化システムにおけるペイントライクなインタフェースの開発, 石田愛; 渡辺知恵美; 城和貴, 我々は現在可視化アプリケーションにおける対話的な問合せツール:Queryballを開発中である.本稿ではQueryballにおいて,ユーザがより積極的に対話的分析を行うためのユーザフレンドリなインタフェースの開発について述べる.Queryballは問合せ条件に該当する(または該当しない)オブジェクトに対する表示方法を持つ半透明の球体である.これまでの仕様ではユーザがQueryballを利用する際,プログラム言語を用いてQueryballの定義を行わなければならなかった.本稿では,そのQueryballの定義を支援するツールの開発について述べる.これによりユーザはより簡単に可視化やQueryballの設定を行うことが可能になる.さらに問合せを適用する領域の指定を,従来の球型だけでなくペン型のツールでも行えるようにし,より直感的に操作ができるようにした., 2004年09月13日, 2004, 92, 1, 4
査読無し, 日本語, 可視化情報学会誌. Suppl., The Visualization Society of Japan, PV-WAVE による極成層圏における大気情報の3次元可視化 : 対話的なビジュアルデータマイニングに向けての試み, 中川 真智子; 渡辺 知恵美; 林田 佐智子; 城 和貴, In recent years visualization technologies are used not only for presentation but also for visual data mining which are techniques for mining new knowledge by visualizing and analyzing the data from multiple view point. In this paper, we considered about visual data mining for atmospheric data, and actually implemented the interactive visualization system which lends to visual data mining using the polar stratospheric data., 2004年07月01日, 24, 1, 53, 54, 10.3154/jvs.24.Supplement1_53
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 可視化システムのための対話的問合せツールにおける問合せ記述支援環境の開発, 石田愛; 渡辺知恵美; 城和貴, 本稿では可視化システムでの対話的問合せツールとして開発しているQueryballの問合せ定義を支援するシステムについて述べる.Queryballは問合せ条件に該当する(または該当しない)オブジェクトに対する表示方法を持つ半透明の球体である.ユーザはQueryballの移動や重ね合わせによって仮想空間の部分問合せを行うことができる.本稿ではまず,可視化パイプラインという記述モデルを用いてQueryballの問合せモデルを定義し,さらにプログラミング言語によるQueryballの定義方法とQueryballについての定義をグラフィカルに閲覧,編集ができるツールを開発した.これによりユーザはQueryballを生成する際に行わなければならない定義を,簡単に行うことが可能となる., 2004年03月01日, 2004, 18, 23, 26
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 可視化システムのための対話的問合せツールの開発と地震波データへの適用, 植村亜以; 渡辺知恵美; 城和貴, 我々は,可視化アプリケーションにおける対話的問合せをQueryballという半透明な球体を用いて行うインタラクションモデルを提案し,プロトタイプシステムの実装を行っている。Queryballは,問合せ条件およびそれに該当する(または該当しない)データの表示方法を持ち,問合せはQueryballの内部にだけ適用される。ユーザは,Queryballを移動したり重ね合わせたりすることで対話的な問合せが可能となる.Queryballの開発を進めていくためには,実際の可視化事例に対して適用し,各事例での適切な利用法を探っていくことが重要である.そこで本稿では,地震波のシミュレーションデータを取り上げ,データの可視化と,Queryballを用いたインタラクションについて述べる。地震波の可現化は,ボリュームレンダリング法を用い,値の分布に基づいて適切な色と透明度の伝達関数を設定した。また,その可視化結果に対して有効と思われるQueryballの利用例について考察し,その一例を適用した。, 2004年03月01日, 2004, 18, 27, 30
査読無し, 日本語, 日本データベース学会letters, 日本データベース学会, Queryball: VRシステムのための対話的な問合せモデル, 渡辺知恵美; 増永良文; 城和貴, 2003年10月, 2, 2, 25, 28
査読無し, 日本語, 情報処理学会研究報告. データベース・システム研究会報告, 社団法人情報処理学会, Queryball : 没入型VRシステムのための対話的な問合せモデル(セッション4C : ユーザインターフェース), 渡辺知恵美; 増永良文; 城和貴, 本稿では,没入型VRシステムにおける新しい問合せの1手法としてQueryballを提案する.Queryballは半透明の球体であり,探索条件および探索条件に該当する(または該当しない)オブジェクトに対する表示方法を持つ.ユーザは生成されたQueryballを仮想世界のオブジェクトに対して重ね合わせることで,Queryballの中にあるオブジェクトに対して問合せを適用させることができる.また,ボールを持って動かす,大きさを変える,複数のQueryballを組み合わせるなどという直感的な操作により試行錯誤的に問合せを行うことができる., 2003年07月16日, 2003, 71, 257, 263
査読無し, 日本語, 電子情報通信学会技術研究報告. DE, データ工学, 社団法人電子情報通信学会, Queryball: 没入型VRシステムのための対話的な問合せモデル(ユーザインタフェース)(「夏のデータベースワークショップ(DBWS2003)」一般), 渡辺知恵美; 増永良文; 城和貴, 本稿では,没入型VRシステムにおける新しい問合せの1手法としてQueryballを提案する.Queryballは半透明の球体であり,探索条件および探索条件に該当する(または該当しない)オブジェクトに対する表示方法を持つ.ユーザは生成されたQueryballを仮想世界のオブジェクトに対して重ね合わせることで, Queryballの中にあるオブジェクトに対して問合せを適用させることができる.また,ボールを持って動かす,大きさを変える,複数のQueryballを組み合わせるなどという直感的な操作により試行錯誤的に問合せを行うことができる., 2003年07月10日, 103, 191, 55, 60
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, Queryball : 没入型VRシステムのための新しい問合せモデル, 渡辺知恵美; 大杉あゆみ; 増永良文; 城和貴, 本稿では、没入型VRシステムにおける新しい問合せの1手法としてQueryballを提案する。Queryballは半透明の球体であり,探索条件および探索条件に該当する(または該当しない)オブジェクトに対する表示方法を持つ。ユーザは生成されたQueryballを仮想世界のオブジェクトに対して重ね合わせることで,Queryballの中にあるオブジェクトに対して問合せを適用させることができる。また、ボールを持って動かす、大きさを変える、複数のQuerybal1を組み合わせるなどという直感的な操作により試行錯誤的に問合せを行うことができる。, 2003年06月24日, 2003, 65, 5, 8
査読無し, その他, 情報処理学会論文誌. コンピューティングシステム, 社団法人情報処理学会, 擬似スペクトル法を用いた乱流場の直接数値シミュレーションの並列化と性能評価, 高田雅美; 山本義暢; 庄野逸; 功刀資彰; 城和貴, 自由界面における乱流を計算機で解析する場合,直接数値シミュレーションと呼ばれる手法により数値解を求める場合がある.この数値解の精度をあげるためには,しばしば擬似スペクトル法が用いられる.本研究では,擬似スペクトル法を用いた乱流解析プログラムを分散メモリ環境対象の並列プログラムに変換し,その性能評価を行った.擬似スペクトル法は,流体の時間発展方程式を解く際,フーリエ変換を用いて波数空間で計算を行う.この際,通常の並列化で使われるstrip miningによる実空間の均等分割によって配列データの分割を行った場合,通信時間と通信オーバヘッドが膨大になる可能性がある.本論文では,この擬似スペクトル法を用いた乱流解析プログラムを並列化するために,通信時間と通信オーバヘッドを考慮し,複数の配列を処理する際に,1つの配列がなるべく少数のプロセッサに割り当てられるように分割し,なおかつ,複数の配列の演算が同時処理できるような方針で実装を行った.提案する並列化手法の有効性を検討するために,並列プログラムをMessage Passing Interface を用いて実装し,安価な分散メモリ環境において性能評価を行った.その結果,現実的な実行時間で解が得られることを確認した., 2003年05月15日, 44, 6, 45, 54
査読無し, 日本語, 情報処理学会論文誌. 数理モデル化と応用, 情報処理学会, Webコミュニティ探索のための量子アルゴリズムの提案と検証, 大音 真由美; 中條 拓伯; 城 和貴, 2003年05月15日, 44, 8, 26, 34
査読無し, 日本語, 電子情報通信学会技術研究報告. NC, ニューロコンピューティング, 社団法人電子情報通信学会, 誤り訂正符合のナイーブ平均場近似, 高田雅美; 庄野逸; 岡田真人; 城和貴, 情報通信の分野において誤り訂正符合は重要な分野であり,近年統計力学的なアプローチを用いて盛んに解析が行われている.本研究では誤り訂正符合の枠組をBayes推定の一種であるMPM(Maximizer of the Posterior Marginals)推定と呼ばれる立場で取り扱い,これにナイーブ平均場近似と呼ばれる近似を適用した場合についての統計力学的解析を行った.ナイーブ平均場近似は,ニューラルネット等の分野において確率的な動作をするバイナリ素子を,連続値を取る決定的な素子に置き換えることによって実現される.この近似は計算時間を短縮する場合などに適用されるが,素子の置き換えによるメリット,デメリットを定量的に評価した研究は非常に少なかった.本研究は近似を行わない場合の解析解と近似を行った場合の解析解を比較し,近似を行なうことのメリットデメリットについて論じ,さらに計算機シミュレーションによる解とを比較検討し解析の正しさを示した., 2003年03月11日, 102, 730, 143, 148
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 大規模シミュレーションデータベース用パラレルI/O制御エージェント, 松岡有希; 松山仁美; 小金山美賀; 上島豊; 城和貴, 大規模シミュレーションを支援する統合管理システムにおいて,現状ではシミュレーション計算によって出力されるデータは,データサーバ上の複数のディスクサイズを考慮することなくセーブされている.そのため,データサーバ上の各ディスクサイズが不均等になり,データサーバの運用が非効率的である,という問題が生じている.そこで我々は,自律的に各ディスク上のデータを監視し,状況に応じてユーザへの問い合わせやデータファイルの移動を行うことによって,ディスクサイズをバランスよく保つデータベース用パラレルI/O制御エージェントのプロトタイプの設計と実装を行う。, 2003年03月03日, 2003, 20, 1, 4
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, パネル討論「数理モデルの今後の展望」, 城和貴, 2003年03月03日, 2003, 20
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 大規模シミュレーションサイクルを統治する知識自動獲得型エージェントプロトタイプの設計と実装(エージェント), 松山仁美; 松岡有希; 小金山美賀; 上島豊; 城和貴, 近年,高性能計算機やネットワークの普及,可視化手法の確立により多種多様な大規模シミュレーション実験が行われている.同時に,一連のシミュレーションサイクルを統合管理するシステムが開発されている.しかし,既存のシステムを効果的に運用するためには,シミュレーションに関するある程度の知識や複雑な操作を必要とするため,ユーザへの負担が大きいのが現状である.そこで,我々はシミュレーションサイクルを自律的に処理するエージェントシステムのプロトタイプを設計し,実装する., 2002年11月28日, 2002, 114, 17, 20
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, ICAを用いた2観測点の地電流データに影響を及ぼす電車ノイズの抽出とその統計的評価(パターン認識), 沢小百合; 小金山美賀; 庄野逸; 長尾年恭; 城和貴, 近年,地震発生直前予測の有効な手法の1つとして,地電流データ中に観測される異常電流の検出が注目されている.この異常電流は地震発生の数週間から数日前に観測されると言われていることから,地震前駆的シグナルと呼ばれている.我が国では電車ノイズの影響で,前駆的シグナルの検出が困難となっている.本研究では,ICA(独立成分分析:Independent Component Analysis)を用いて,独立信号であると考えられる電車ノイズの自動分離を試みている.これまでの実験で,長野県松代観測点の地電流データにICAを適用すると電車ノイズを分離できることが確認されている,本稿では,同じ電車ノイズの影響を受けると考えられる2観測点の地電流データにそれぞれICAを適用し,両観測点から同じ電車ノイズが抽出できるかどうかの実験と実験結果の統計的な評価を行った., 2002年11月28日, 2002, 114, 95, 98
査読無し, その他, 情報処理学会論文誌. 数理モデル化と応用, 社団法人情報処理学会, 量子コンピュータ実現に向けた量子命令セットについて(<特集>量子計算), 大音真由美; 中條拓伯; 高田司郎; 城和貴, 量子コンピュータ開発に向けた研究は始まったばかりで,量子素子・デバイスと,量子チューリング機械上のアルゴリズムの2つの興味深い課題が研究されている.これらの研究は非常に重要な基礎研究である.しかし,我々は,量子コンピュータを実現するためには,量子コンピュータ.アーキテクチャ,および,そのシステム・ソフトウェア側からの研究が不可欠であると確信している.そこで,本論文では,量子コンピュータ・アーキテクチャ構築の準備として,量子コンピュータ実現に向けた量子命令セットを提案する.具体的には,まず,現在までに提案されている5つの量子アルゴリズムに共通な量子基本操作を抽出する.次に,それら量子基本操作を実現するために,量子ユニットモデルとそのモデル上で実行する量子命令セットを提案する.さらに,この命令セットを上記5つの量子アルゴリズムのコーディングに適用することで,提案する量子命令セットの有効性を示す.最後に,今後の課題である量子コンピュータ・アーキテクチャの枠組みの一例を示す., 2002年09月15日, 43, 7, 19, 28
査読無し, その他, 情報処理学会論文誌. 数理モデル化と応用, 社団法人情報処理学会, ICAを用いた地電流データからの電車ノイズと地震前兆シグナルの分離, 小金山美賀; 庄野逸; 長尾年恭; 城和貴, 地電流データ中には,地震前兆シグナルと呼ばれている異常電流が観測されることがある.近年,地電流データからの地震前兆シグナルの検出は,短期地震予知の有効な手法の1つとして注目されている.しかし,我が国では地電流データ中の多くに観測されている電車ノイズの影響で,地震前兆シグナルの検出が困難となっている.本研究では信号分離技術であるICA(独立成分分析:Independent Component Analysis)を用いて,独立信号であると考えられる電車ノイズや地震前兆シグナルの自動分離を試みている.本論文では,地電流データにICAを適用し,その結果と評価について報告する., 2002年09月15日, 43, 7, 92, 104
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, ループ内依存関係の視覚的パターン化の試み, 岩坂麻実; 山口智美; 庄野逸; 笹倉万里子; 城和貴, 並列計算機の普及に伴い逐次プログラムを並列プログラムに書き換える必要があるが,現在のところ,プログラマに対して高度な技術や経験が要求される.本研究では,並列支援視覚化システムNaraViewを用いて,与えられたプログラムのデータ依存関係を視覚的パターン化し,適切な変換手法に対応づけることを考案する.本稿では主に,並列化コンパイラで一般に使用されているループ変換手法について取り上げ,具体的に視覚化されたプログラム中のループのデータ依存状態から有効なループ変換手法を検証する., 2002年03月04日, 2002, 19, 41, 44
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 自動並列化コンパイラPROMIS用ユニモジュラ変換の設計と実装, 石内寿子; 山口智美; 庄野逸; 城和貴, 自動並列化コンパイラにおいて、並列度を増加させるための最適化手法として様々なループ変換法が提案されてきた。しかし、これらの方法は個々に開発されてきたため、それぞれ適応条件や効果が異なる。そのため、コンパイラに複数のループ変換法を実装しても、その中からどの変換が最適であるかを決定することが難しい問題である。一方、ユニモジュラ変換という、標準的なループ変換のいくつかの組み合わせによる変換を一度に行える変換がある。この一度の変換で、並列化のための最適解を求めることができるのである。このユニモジュラ変換を、イリノイ大学で開発された自動並列化コンパイラPROMISに付け加えることが本研究の目的である。, 2002年03月04日, 2002, 19, 45, 48
査読無し, その他, 情報処理学会第64回全国大会予稿集, 特別トラック 4D-04, 汎用量子コンピュータのモデルとシミュレータの構想, 古屋 良二朗; 中條 拓伯; 城 和貴, 2002年, 249, 252
査読無し, その他, 第43回プログラミング・シンポジウム 報告集, 量子コンピュータの命令セットアーキテクチャの一提案とそのシミュレータの構想, 中條 拓伯; 古屋 良二郎; 大音 真由美; 城和貴, 2002年, 173, 184
査読無し, その他, 情報処理学会論文誌. 数理モデル化と応用, 社団法人情報処理学会, 「情報処理学会論文誌 : 数理モデル化と応用」の発行にあたって, 城和貴, 2001年12月15日, 42, 14
査読無し, 日本語, 情報処理学会論文誌. 数理モデル化と応用, 社団法人情報処理学会, ニューラルネットを用いた地電流データからの電車ノイズ除去, 小金山美賀; 長尾年恭; 城和貴, 短期地震予知の手法として, 近年, VAN法という観測方法で採取された地電流データから地震前兆シグナルを検出し, 短期的な地震予知を行う方法が注目されている.しかし, 我が国では, 地電流データ中の多くの占める電車ノイズの影響で, 地電流データから地震前兆シグナルを検出することが困難になっている.そこで, 我々は, ニューラルネットを利用して, 地電流データから地震前兆シグナルを自動的に検出し, 実用的かつ信頼性の高い短期地震予知を行う研究に着手している.本論文では, その第1ステップとして, 地震前兆シグナルを検出する際に問題となっている電車ノイズを除去するニューラルネットの構築法とその評価について報告する., 2001年12月15日, 42, 14, 124, 133
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 短期地震予知を目的とした地電流データへのICAの適用と評価, 小金山美賀; 庄野逸; 長尾年恭; 城和貴, 地電流データ中には, 地震前兆シグナルと呼ばれている異常電流が観測されることがある.近年, 地電流データからの地震前兆シグナルの検出は, 短期地震予知の有効な手法の1つとして注目されている.しかし, 我が国では地電流データ中の多くに観測されている電車ノイズの影響で, 地震前兆シグナルの検出が困難となっている.本研究ではICA(独立成分分析:Independent Component Analysis)を用いて, 独立信号であると考えられる電車ノイズや地震前兆シグナルの自動分離を試みている.これまでに, 長野県・松代観測点の地電流データにICAを適用すると, 電車ノイズを分離できることが実験によって確認されている.本稿では, 他の観測点でも電車ノイズを分離可能であるかを検証するために, 松代観測点よりも電車の本数が多い福井県・笹谷観測点の地電流データにICAを適用し, その結果と評価について報告する., 2001年11月19日, 2001, 110, 9, 12
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, シミュレーション試行を前提とした量子コンピュータ・アーキテクチャ, 古屋良二郎; 大音真由美; 城和貴; 中條拓伯, 近年, 量子コンピューティングに関する研究が活発になっている.研究の潮流としては, 効率的量子アルゴリズムに関する研究と実験物理学の領域において量子デバイスの実現を目指す研究に大別される.我々は, 将来の量子コンピュータシステムの実現を想定とし, 汎用的な量子コンピュータの抽象モデルを探求している.本論文では, シミュレーション試行を前提とした量子コンピュータ・アーキテキチャを提案し, 量子命令セットおよび内部アーキテキチャを定義した.この抽象モデルは, 種々の量子命令実行のシミュレーションを行う際の基礎設計部分となるものである., 2001年11月19日, 2001, 110, 19, 22
査読無し, 日本語, 情報処理学会研究報告. 計算機アーキテクチャ研究会報告, 社団法人情報処理学会, 自動並列化コンパイラPROMIS-NWUの概要, 山口智美; 石内寿子; 岩坂麻実; 羽田昌代; 庄野逸; 城和貴, コンパイラは高コスト・ソフトウェア開発の代表のように扱われているが, 近年様々な技術革新により, その開発コストの削減が図られている.例えばオブジェクト指向技術の導入により, コストは数分の一に減少する.このようなソフトウェア工学的なアプローチの他に, コンパイラの構成要素を再利用する形での開発の低コスト化が期待されている.本研究グループでは, イリノイ大学で開発された自動並列化コンパイラPROMISの中間表現を入れ替えることで, 異なるコンセプトに基づくコンパイラを低コストで開発している.本稿では開発中の自動並列化コンパイラPROMIS-NWUの概要について報告する., 2001年07月25日, 2001, 76, 79, 84
査読無し, 日本語, 電子情報通信学会技術研究報告. NC, ニューロコンピューティング, 社団法人電子情報通信学会, ICAを用いた地電流データからの電車ノイズおよび地震前兆シグナルの分離, 小金山美賀; 庄野逸; 杉田望代; 長尾年恭; 城和貴, 近年, 短期地震予知の手法として, VAN法という観測方法を用いて採取された地電流データから地震前兆シグナルを検出し, 短期的な地震予知を行う方法が注目されている.しかし, 我が国では地電流データ中の多くに観測されている電車ノイズの影響で, 地震前兆シグナルの検出が困難になっている.本研究では, ICA(独立成分分析:Independent Component Analysis)を用いて独立なソースと考えられる電車ノイズや地震前兆シグナルを地電流データから分離し, 手作業では困難である地震前兆シグナルの検出を自動的に行うことを試みている.本稿では, 長野県・松代観測点のいくつかの地電流データをICAに適用した.その結果, 電車ノイズや地震前兆シグナルと考えられるソースが出力されていることを確認することができた., 2001年07月20日, 101, 238, 79, 85
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, LVQを用いた地電流データ解析II, 福田京子; 小金山美賀; 長尾年恭; 城和貴, 我々の研究グループでは、VAN法を用いて観測した地電流データをニューラルネットにより自動的に解析し、短期地震予知を行う研究に着手している。既にバックプロパゲーション法による電車ノイズと地震前兆シグナルの分離に成功している。前回、バックプロパゲーションの課題である膨大な計算時間短縮の解決策として、LVQによる地電流データ解析を長野県松代地点のデータを用いて提案した。本稿では、前回の問題点であった電車ノイズの影響がほとんどない新島のデータを用いてLVQの構築と、その検証について報告する。, 2001年03月15日, 2001, 27, 33, 36
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 自動並列化コンパイラの統一的中間表現に対する基本APIの実装, 山口智美; 城和貴, 自動並列化コンパイラの開発に際し,様々な問題を考慮しそれらの問題を克服し得るコンパイラを実装することは非常に難しく,開発期間も長くなることが一般的である.この問題に対して,複数の中間表現を統合した統一的中間表現(UIR)を定義し,最適化等の変換手法をそのインタフェイスを通じて行うことが提唱されている.本稿ではUIRインタフェイス・ライブラリの整備とそれを使ったループ変換手法の実装を通じて,UIRインタフェイスを定義することの意義について議論を行う., 2001年03月15日, 2001, 27, 37, 40
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, DirectXを用いたNaraViewの高速化, 柴山智子; 城和貴, NaraViewは与えられたプログラムから必要な情報を抽出し、3次元空間上に視覚化する。その役割は人間の優れた視覚認識能力による判別を通しユーザー自身にプログラム構造を把握させることでプログラムの並列化を支援することである。本稿ではNaraViewの現状を踏まえた上で、DirectXを用いた再実装の試みについて述べる。, 2001年03月15日, 2001, 27, 41, 44
査読無し, 日本語, 情報処理学会研究報告. 計算機アーキテクチャ研究会報告, 社団法人情報処理学会, 乱流シミュレーションの並列化と評価, 高田雅美; 山本義暢; 功刀資彰; 城和貴, 本稿では, 自由表面乱流場における直接数値シミュレーションの逐次プログラムを, 並列化ライブラリMessage Passing Interface(MPI)を用いて並列プログラムに変換するための並列化手法について報告する.4ならびに8プロセッサ対応の並列プログラムを実装し, それらの実行時間が, 逐次プログラムと比較して, それぞれ約1/4, 1/8になることを, 実験によって示す., 2001年01月26日, 2001, 10, 1, 6
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, LVQを用いた地電流データ解析, 福田京子; 小金山美賀; 長尾年恭; 城和貴, 我々の研究グループでは、VAN法を用いて観測した地電流データをニューラルネットにより自動的に解析し、短期地震予知を行う研究に着手している。既にバックプロパゲーション法による電車ノイズと地震前兆シグナルの分離に成功したが、本稿では、バックプロパゲーションの課題である膨大な計算時間短縮の解決策として、LVQによる地電流データ解析を提案の構築と、その検証について報告する。, 2000年11月21日, 2000, 108, 25, 28
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, NaraViewのParafrase-2との連携における効果的な並列化支援, 羽田昌代; 笹倉万里子; 長嶋雲兵; 國枝義敏; 城和貴, 並列計算機を効果的に使用するために, アプリケーションの並列化を自動で行う自動並列化コンパイラが開発されているが, これを用いても正確で効果的な並列化を行うことは極めて難しい.このため自動並列化コンパイラには, その使用を支援するツールが開発されていることが多い.我々が, これまでに開発している並列化支援ツールNaraViewは並列化コンパイラParafrase-2から得られる中間表現の情報を抽出し, 与えられたプログラムの視覚化を行う.本稿では実アプリケーションの並列化を通してNaraViewのより有効な使用について述べる., 2000年06月26日, 2000, 60, 17, 20
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 短期地震予知のためのニューラルネットを用いた地電流データからの電車ノイズ除去, 小金山美賀; 杉田望代; 長尾年恭; 城和貴, 我々は, VAN法を用いて観測される地電流データから, 地震前兆シグナルを計算機により自動的に検出し, より実用的かつ信頼性の高い短期地震予知を行う研究に着手している.本稿では, 地震前兆シグナルを検出する際に問題となる電車ノイズの除去が, ニューラルネットを用いることによって, 可能かどうかの実験を行い, その結果について報告する., 2000年06月26日, 2000, 60, 29, 32
査読無し, 日本語, 情報処理学会研究報告. [ハイパフォーマンスコンピューティング], 社団法人情報処理学会, 2000-HPC-81-8 NaraViewを利用した実アプリケーションの並列化事例, 羽田昌代; 笹倉万里子; 城和貴, 並列計算機を効果的に使用するためには, これまで逐次計算機で使用されてきたアプリケーションを並列プログラムに再構築する必要があるが, その作業は煩雑で容易ではない.アプリケーションの並列化を自動で行う自動並列化コンパイラが開発されているが, これを用いても正確で効果的な並列化を行うことは極めて難しい.このため自動並列化コンパイラには、その使用を支援するツールが開発されていることが多い.我々はこれまでにNaraViewという並列化支援ツールを開発している.NaraViewは並列化コンパイラ Parafrase-2から得られる中間表現の情報を抽出し, 与えられたプログラムの視覚化を行う.本稿では実アプリケーションの並列化を通してこのNaraViewの有用性を評価する., 2000年06月02日, 2000, 57, 39, 44
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, ニューラルネットワークによるプログラム分割アルゴリズムの改善, 齊藤哲哉; 高田雅美; 城和貴; 國枝義敏; 福田晃, 分枝限定法を用いて最適なプログラム分割を求めるGirkarのプログラム分割アルゴリズムでは, 限定操作に用いられる有効な下限値が提案されておらず, また, そのような下限値を求めることは困難であるため, 大規模な問題に対して適用しづらかった.本稿では, 限定操作を行う際に利用する下限値の予測にニューラルネットワークを導入することを提案し, 最適ではない可能性があるが, 実用的に問題のない程度の近似解を得ることを目指す., 2000年02月07日, 2000, 16, 21, 24
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 変数アクセスコストを引力としてとらえたデータ分割手法とその視覚化, 杉田望代; 大音真由美; 笹倉万里子; 城和貴, 分散メモリ型環境において効率よく並列処理するには、"複数のメモリにデータをどのように分割・配置するか"というデータ分割問題を考えなければならない。本稿では、分散メモリ環境におけるデータ分割問題と自動並列化コンパイラの内部表現について述べた後、物理モデルである引力を応用したデータとタスクのクラスタ化支援とその視覚化の説明を行い、本視覚化を用いたデータ分割手法を提案する。さらに、そのプロトタイプを用いて、本手法の有効性を述べる。, 2000年02月07日, 2000, 16, 33, 36
査読無し, 日本語, 情報処理学会研究報告. 計算機アーキテクチャ研究会報告, 社団法人情報処理学会, コンパイラ制御可能なCOMA環境を実現するソフトウェアDSM Laurasiaの実装, 上原哲太郎; 齋藤彰一; 城和貴; 國枝義敏, 自動並列化コンパイラの作成に際し、並列計算機アーキテクチャのノード間通信機構の差異を、コンパイラから制御可能なCOMA(Cache Only Memory Architecture)環境の実現によって隠蔽する方式を提案している。本論文では、同方式によるコンパイラをUNIXベースのワークステーション(WS)クラスタ上に実現するために作成したソフトウェアDSM(分散共有メモリ)の実現について述べる。共有メモリの弱い整合性制御モデルとしてEntry Consistencyの実現機構を備える。100BASE-TX使用時、4〜6MB / sのページ転送性能を達成した。, 2000年01月07日, 2000, 1, 49, 54
査読無し, その他, 情報処理学会 「第7回MPSシンポジウム」 新しい計算パラダイム ~量子/分子コンピュータ最前線~ 論文集, 汎用量子コンピュータ・アーキテクチャの構想, 大音 真由美; 中條 拓伯; 城 和貴, 2000年, Vol.2000, No.16, 77, 80
査読無し, その他, 情報処理学会論文誌. 数理モデル化と応用, 社団法人情報処理学会, リリース・コンシステンシ・モデルとその実現の形式的仕様記述について, 高田司郎; 田口研治; 城和貴; 福田晃, 我々は,形式的仕様記述言語Zの表記法とプロセス代数value-passing CCSを統合した形式手法を用いて,分散共有メモリシステムの振る舞いを定義したメモリ・コンシステンシ・モデルとそれらの実現の形式的な仕様記述と検証の研究を行っている.メモリ・コンシステンシ・モデルは,ストア命令やロード命令などに諸々のプログラム順序を定義してメモリアクセスを制約するものと,いつどのようにこれら命令の同期を取るべきかというプログラマの指定によりメモリアクセスを制約するものに大別される.我々は,以前,前者の代表としてコーザル・メモリ・コンシステンシ・モデルを取り上げ,このモデルと実現の形式的な仕様記述と検証に対するこの形式手法の有効性を報告した.そこで,本論文では,後者の代表としてリリース・コンシステンシ・モデルを取り上げ,このモデルと実現の形式的な仕様記述と検証を行なう.これら代表的な二つのモデルと実現の形式的仕様記述と検証を示すことで,この形式手法のメモリ・コンシステンシ・モデルに対する有効性を確認した., 1999年12月15日, 40, 9, 1, 17
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 地電流観測データのニューラルネットワークによる解析, 小金山美賀; 長尾年恭; 城和貴, 我々は,VAN法を用いて観測した地電流データをニューラルネットワークにより自動的に解析し,短期地震予知を行う研究に着手している.本稿では,予備実験として,地電流データから地震前兆波を検出しやすくするために,地電流データに含まれている電車のノイズ波を学習,識別するニューラルネットワークを構築し,その検証について報告する.さらに,元データにfftを用いたフィルタリングを行うことによって,学習の高速化が可能であるかどうかを検証する., 1999年11月25日, 99, 96, 1, 4
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 実用的な近似解を与えるプログラム分割アルゴリズム, 齋藤哲哉; 高田雅美; 城和貴; 國枝義敏; 福田晃, 本論文では,Girkarのプログラム分割アルゴリズムに対してヒューリスティックスを導入し,最適な解が得られる保証はないが,実用的に問題のない程度の近似解が得られることを示す.我々が提案するヒューリスティックスは,対象となるタスク・グラフの枝のリストの並びに注目し,ある枝の位置を基点とし,そこから選択した枝までの距離の平均値を切り上げた値と最大値を評価値として利用する.この評価値によって,いくつ先まで分枝操作をするかを決定し,得られた暫定値を下限値として限定操作を行う.実験の結果,十分に実用的な近似解が得られることがわかった., 1999年11月25日, 99, 96, 29, 32
査読無し, その他, 電子情報通信学会技術研究報告. IE, 画像工学, 社団法人電子情報通信学会, 自動並列化コンパイラ内部情報の三次元視覚化システム, 城和貴; 杉田望代, 本稿では、自動並列化コンパイラにおける統一的中間表現について概説し、その一例であるデータ分割グラフ(DPG)の特徴を紹介する。そして、一般的な視覚化の有効性を述べた後、DPGの視覚化を提案する。DPGのデータアクセス情報は、変数単位でデータが表現され、アクセスの種類も分けられており、詳細な内容を含んでいる。そこで、最適なデータのメモリ配置の支援を目的とする、DPGの変数アクセス情報の視覚化の提案を行う。さらに、そのプロトタイプ実装について報告した後、具体例をもって本提案の有効性を検証する。, 1999年10月28日, 99, 400, 59, 66
査読無し, その他, 電子情報通信学会技術研究報告. DSP, ディジタル信号処理, 社団法人電子情報通信学会, 自動並列化コンパイラ内部情報の三次元視覚化システム, 城和貴; 杉田望代, 本稿では、自動並列化コンパイラにおける統一的中間表現について概説し、その一例であるデータ分害1jグラフ(DPG)の特徴を紹介する。そして、一般的な視覚化の有効性を述べた後、DPGの視覚化を提案する。DPGのデータアクセス情報は、変数単位でデータが表現され、アクセスの種類も分けられており、詳細な内容を含んでいる。そこで、最適なデータのメモリ配置の支援を目的とする、DPGの変数アクセス情報の視覚化の提案を行う。さらに、そのプロトタイプ実装について報告した後、具体例をもって本提案の有効性を検証する。, 1999年10月28日, 99, 396, 59, 66
査読無し, その他, 電子情報通信学会技術研究報告. ICD, 集積回路, 電子情報通信学会, 自動並列化コンパイラ内部情報の三次元視覚化システム, 城和貴; 杉田望代, 本稿では、自動並列化コンパイラにおける統一的中間表現について概説し、その一例であるデータ分割グ簗霞宵門翼;・4義1理胃z胃本壜53エソ品畠容を含んでいる。そこで、最適なデータのメモリ配置の支援を目的とする、DPGの変数アクセス情報の視覚化の提案を行う。さらに、そのプロトタイプ実装について報告した後、具体例をもって本提案の有効性を検証する。, 1999年10月28日, 99, 398, 59, 66
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 周期性を有する配列参照パターンの解析手法について, 小崎光倫; 中西恒夫; 城和貴; 福田晃, 分散メモリ型並列計算システムにおいては,通信オーバーヘッドの問題から,各プロセッサが頻繁に参照するデータをそのローカルメモリに局所的に配置するデータ配置最適化が極めて重要である.しかしながら,科学技術計算アプリケーションでは,ループ中で複雑な配列参照が行われ,その参照パターンを明確に把握するのは非常に難しく,コンパイル時に十分にデータ配置最適化を行うことが困難である.本稿では,実行時に配列参照に関するログをとることにより,そのログから配列参照の周期性の解析を試みる.配列参照周期が明らかになれば,データ配置最適化などの最適化処理への利用が期待される., 1999年09月21日, 99, 76, 1, 4
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, データ分割グラフの3次元視覚化, 杉田望代; 笹倉万里子; 城和貴, 本稿では、自動並列化コンパイラにおける統一的中間表現について概説し、その一例であるデータ分割グラフ(DPG)の特徴を紹介する。そして、一般的な視覚化の有効性を述べた後、DPGの視覚化を提案する。DPGのデータアクセス情報は、変数単位でデータが表現され、アクセスの種類も分けられており、詳細な内容を含んでいる。そこで、最適なデータのメモリ配置の支援を目的とする、DPGの変数アクセス情報の視覚化の提案を行う。さらに、そのプロトタイプ実装について報告した後、具体例をもって本提案の有効性を検証する。, 1999年09月21日, 99, 76, 21, 24
査読無し, その他, 情報処理学会研究報告. 計算機アーキテクチャ研究会報告, 社団法人情報処理学会, 分枝限定法を用いたプログラム分割の下界に対する考察, 高田雅美; 齊藤哲哉; 中西恒夫; 城和貴, 本稿では,非循環であるタスクグラフを対象としたGirkarのプログラム分割アルゴリズムを見直し,その改良として下限値の1つを提案し,実験によって有効性を示した.下限値を用いた探索の制限により多くの場合,探索時間は短くなり,解についても最適解が返された.稀に解が最適とならない場合もあるが,最適解からの誤差は小さいので,解は近似である.また,Girkarのアルゴリズムではメモリ不足となり実行不可能となるタスクグラフについても,実行結果を得ることを可能とした., 1999年08月02日, 99, 67, 79, 84
査読無し, その他, 情報処理学会研究報告. 計算機アーキテクチャ研究会報告, 社団法人情報処理学会, 異なるコンパイラ中間表現の相互変換に関する考察, 曽山典子; 神戸和子; 城和貴; 加古富志雄, 中間表現の研究は永い年月をかけて多くの研究者によって研究されてきた.中間表現の基本構成は共通の認識によって構築されているにもかかわらず,多くのコンパイラでは独自の中間表現を構築し,その実装に多くの時間を費やしている.近年,コンパイラ技術の融合を図るべく統一的中間表現が提案されているが,すでに開発されたコンパイラ技術との融合は困難である.本稿では,すでに実装された異なる中間表現を比較し,それらの相互変換の可能性を調べ,その結果,2つの異なる中間表現は実装上の細部では困難な部分もあるが,概ね可能であることがわかった.同時に各構造体の構成成分がおおよそ同じであることがわかった., 1999年08月02日, 99, 67, 85, 90
査読無し, その他, 情報処理学会研究報告. 計算機アーキテクチャ研究会報告, 社団法人情報処理学会, 配列参照パターンによるプログラム並列化・最適化支援, 奥田宗継; 中西恒夫; 笹倉万里子; 城和貴; 福田晃, 「控え目な並列化.最適化」を行う自動並列化コンパイラは,しばしばプログラムの並列化・最適化に失敗し,出力されたプログラムに並列化・最適化の余地が残っていることが多い.プログラマが並列化・最適化に必要な情報を取得し,並列化・最適化を手動で適用するインタラクティブコンパイル環境の実現が望まれる.本稿では,並列化支援視覚化システムNaraViewを用いてループにおける配列参照パターンを視覚化し,並列化・最適化を施す方法に関する議論を行う., 1999年08月02日, 99, 67, 145, 150
査読無し, その他, 情報処理学会研究報告. 計算機アーキテクチャ研究会報告, 社団法人情報処理学会, 自動並列化コンパイラの統一的中間表現とインターフェースを用いたコード変換の実装, 羽田昌代; 神戸和子; 中西恒夫; 城和貴, 自動並列化コンパイラで逐次プログラムを並列化する時の手法としてコード変換がある.従来のコード変換はそれぞれの対応する中間表現上で実装が行われており適用順序が任意ではない等の不都合があった.これらの中間表現を統一し,その一つの中間表現に対して変換を行うことが提案されている.本稿では,この統一的中間表現とそのインターフェースを用いて,コード変換手法の一つであるインライン展開を実装し,その結果を検討し統一的中間表現を用いることの意義について議論を行う., 1999年08月02日, 99, 67, 151, 156
査読無し, その他, 情報処理学会研究報告. [ハイパフォーマンスコンピューティング], 社団法人情報処理学会, WindowsNT上でのクロス開発環境を目指す並列化支援ツールの開発, 水原隆道; 平尾智也; 齊藤哲哉; 笹倉万里子; 城和貴; 國枝義敏, 現在の自動並列化コンパイラでは,逐次プログラムを効率良く完全に自動並列化することは困難である.より効率の良い並列化を行うためには,利用者がどの部分をどのように並列化するかを,明示的に指示する必要がある.並列化支援可視化システムNaraViewでは,並列化の度合いを含むプログラム構造,データの依存関係などの情報を視覚化して利用者に提示することが可能である.我々は,NaraViewをWindowsNT上に移植し,従来の機能に加えてソースプログラムとの連係を図る.さらに,ユーザーインターフェイスの改善や,自動並列化コンパイラParafrase-2との連係を強化したシステムを開発し,評価を行う., 1999年08月02日, 99, 66, 35, 40
査読無し, その他, 情報処理学会研究報告. [ハイパフォーマンスコンピューティング], 社団法人情報処理学会, タンパク質分子構造を例とする高性能計算結果の可視化システムの試作, 平尾智也; 水原隆道; 齊藤哲哉; 安藤誠; 秋山泰; 城和貴; 國枝義敏, 高性能計算を通じて生成される膨大な計算結果の可視化手法の研究の一環として,我々はタンパク質分子構造の高度な可視化を行なう"ProteinView"システムを作成した.ProteinViewはWindows95/98/NT4.0上で動作し,タンパク質分子の立体構造を三次元の仮想空間内に効率よく表示するとともに,ジョイスティックを用いて操作者の視点位置を滑らかに移動させることができる.従来のシステムの多くが,分子構造全体を外部から眺めることに重点を置いてきたのに対し,本手法では巨大分子の内部空間における分子の局所構造の詳細な検討が容易に行なえる.本稿では,ProteinViewの設計上の特徴や,実装手法などを述べる., 1999年08月02日, 99, 66, 125, 130
査読無し, 日本語, 電子情報通信学会技術研究報告. SS, ソフトウェアサイエンス, 社団法人電子情報通信学会, 分散共有メモリの形式的仕様記述について, 高田司郎; 田口研治; 城和貴; 福田晃, 本稿では, ZとValue-passing CCSを統合した形式手法を用いて,分散共有メモリシステムの振る舞いを定義したメモリ・コンシステンシ・モデルとそれらの実現の形式的な仕様記述と検証を行う. メモリ・コンシステンシ・モデルは, ストア命令やロード命令などに諸々のプログラム順序を定義してメモリアクセスを制約するものと, いつどのようにこれら命令の同期を取るべきかというプログラマの指定によりメモリアクセスを制約するものに大別される. 本稿では, 前者の代表としてコーザル・メモリ・コンシステンシ・モデルを, 後者の代表としてリリース・コンシステンシ・モデルを取り上げる. これらの代表的な二つのモデルとそれらの実現の形式的仕様記述と検証を示すことで, この形式手法の分散共有メモリに対する有効性を確認した., 1999年03月19日, 98, 676, 41, 48
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 剰余区間演算 : データフロー解析のための数学的ツール, 中西恒夫; 城和貴; PolychronopoulosConstantine D; 福田晃, 実数上の区間に対する演算体系として区間演算がある. 本来, 区間演算は浮動小数点計算の丸め誤差解析を行うために考案されたものであるが, プログラム解析にも応用されている. しかしながら, プログラムのループの指標あるいは配列変数の添字等に用いられるのは専ら整数であり, 従来の実数上の区間ではプログラム解析に必要な情報を十分に表現できない. 本稿では, 区間演算を拡張し, 実数上の区間に含まれる整数の集合に対する演算体系, 剰余区間演算を定義し, その数学的性質を明らかにする. さらに, 自動ベクトル化/並列化コンパイラにおいて行われるデータフロー解析への剰余区間演算の応用を検討する., 1999年02月18日, 99, 15, 37, 42
査読無し, その他, 情報処理学会論文誌. 数理モデル化と応用, 社団法人情報処理学会, 変数オリエンティッドなデータ依存関係モデルの提案, 笹倉万里子; 城和貴; 國枝義敏; 荒木啓二郎, プログラムを特徴づける性質の一つとしてデータ依存関係がある. これまで, データ依存関係は, ソースプログラムの文の間の関係としてコンパイラの内部表現で表されていた. しかし近年, ユーザによるプログラムの理解, 最適化や並列化の観点から, ユーザによるデータ依存関係の理解が重要性を増している. そこで本論文では, ユーザへのデータ依存関係の提示を主目的として, 変数を中心としてデータ依存関係を記述するモデルを提案し, それを用いたデータ依存関係の視覚化法について述べる. また, 提案するデータ依存関係の表現法と従来の表現法との対応について述べる., 1999年02月15日, 40, 2, 45, 54
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, リリース・コンシステンシ・モデルとその実現の形式的仕様記述について, 高田司郎; 田口研治; 城和貴; 福田晃, 我々は, 分散共有メモリシステムの振る舞いを定義したメモリ・コンシステンシ・モデルとその実現の形式的な仕様記述とその検証の研究を行っている.我々は, 仕様記述言語Zとvalue-passing CCSを統合した形式的技法をコーザル・メモリ・コンシステンシ・モデルに適用した結果, そのモデルと実現の形式的仕様記述, および, その実現の検証がこの技法で一貫して行えることを確認した.そこで, 本稿では, リリース・コンシステンシ・モデルに同様の形式的技法を適用して, そのモデルとその同期に関する実現の形式的仕様記述を行った., 1998年11月26日, 98, 105, 81, 86
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, A Combined Data and Program Partitioning Algorithm for Distributed Memory Multiprocessors, 中西恒夫; 城和貴; ConstantineD.Polychronopoulos; 福田晃, In this paper we propose an algorithm to perform data partitioning and program partitioning simultaneously on the Data Partitioning Graph, an intermediate representation for parallelizing compilers. Conventional and, therefore, conservative parallelizing compilers usually activate program partitioning prior to data partitioning. However, on distributed memory multiprocessors, since communication costs change depending on a data partitioning and distribution decision, it is quite difficult to partition a program effectively with consideration of data partitioning. The proposed algorithm reso..., 1998年11月26日, 98, 105, 87, 92
査読無し, その他, 情報処理, 社団法人情報処理学会, PDPTA'98, 城和貴, 1998年09月15日, 39, 9, 940, 941
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 変数オリエンティッドなデータ依存関係モデルの提案, 笹倉万里子; 城和貴; 國枝義敏; 荒木啓二郎, プログラムを特徴づける性質の一つとしてデータ依存関係がある.これまで, データ依存関係は, ソースプログラムの文の間の関係としてコンパイラの内部表現で表されていたが, ユーザに提示されることはなかった.しかし近年, ユーザによるプログラムの理解, 最適化や並列化の観点から, ユーザによるデータ依存関係の理解が重要性を増している.そこで本論文では, ユーザへのデータ依存関係の提示を主目的として, 変数を中心としてデータ依存関係記述するモデルを提案し, それを用いたデータ依存関係の視覚化法について述べる.また, 提案するデータ依存関係の表現法と従来の表現法との対応について述べる., 1998年07月24日, 98, 67, 31, 36
査読無し, 日本語, 情報処理学会研究報告. PRO, [プログラミング], 社団法人情報処理学会, 統一的中間表現を用いた自動並列化コンパイラの実装 : ソースコードから統一的中間表現への変換, 田中栄治; 中西恒夫; 城和貴; 山下雅史, 従来の自動並列化コンパイラで一般的に用いられる中間表現は, タスク間の依存関係のみを表現するのみが多く, データの分割配置ならびに転送の最適化手法において重要なデータフローの情報が明示的に表現されていない. そのためこれらの最適化手法は中間表現から遊離した形で処理されることが多く, これは一貫性のとれた並列化・最適化を難しくしている. 本研究で実装している自動並列化コンパイラ Narafrase では, データフロー情報を明示的に表現する中間表現 DPG を統一的中間表現, すなわち全並列化・最適化手法の共通の中間表現とし, 一貫性のとれた並列化・最適化の実現を図る. 本稿では,ソースコードから DPG への変換パスの実装について報告する., 1998年01月23日, 98, 7, 19, 24
査読無し, 日本語, 電子情報通信学会論文誌. D-I, 情報・システム, I-コンピュータ, 社団法人電子情報通信学会, ループ最小並列実行時間算出の一手法, 中西恒夫; 城和貴; ポリクロノポーラスコンスタンチンD; 荒木啓二郎; 福田晃, 本論文では, ループの最小並列実行時間を算出する一手法を提案する. ループを完全に展開したコードのタスクグラフの節点, 枝のそれぞれに, 当該タスクの実行時間, 当該依存に関与する通信の時間をコストとして与えた場合, クリティカルパス -すなわちパスを構成する節点, 枝のコストの合計が最大となるパス- のコストはループの最小並列実行時間を与える. しかしながら, ループを完全に展開することは, 計算時間ならびに消費記憶量の点で実用上好ましくない. 本論文で述べる手法は, 問題を整数計画問題として定式化し, ループを一切展開することなく, このクリティカルパスのコストを算出する. また本論文では, 定式化された整数計画問題を解く分枝限定法ベースのアルゴリズムを実装する. 実装アルゴリズムは, ループを展開したコードのタスクグラフ上のクリティカルパスを与える. Livermoreベンチマークカーネルを用いた評価の結果, 実装アルゴリズムの計算時間は十分実用的であり, かつループの大きさ(繰返し数)に依存しないことが確認された., 1997年09月25日, 80, 9, 763, 773
査読無し, 日本語, 全国大会講演論文集, 社団法人情報処理学会, オブジェクト指向に基づくスカラ拡張の適用条件のクラス化, 大森洋一; 城和貴; 福田晃, マルチプロセッサを対象とした並列化コンパイラは, 従来のコンパイラと比べて複雑なデータ構造を使用するため, 保守性が犠牲になりやすい。その反面, マルチプロセッサには多様なアーキテクチャが存在するので, コンパイラ内部の柔軟性を向上させるのが重要な課題となっている。そこで我々は, オブジェクト指向に基づく並列化コンパイラの設計と, それに基づく試作を行なっている。本稿では, オブジェクト指向に基づくコンパイラ設計の拡張性を検証するために, ループ・プライベイタイゼーション(loop privatization)と呼ばれる並列化手法をクラスとして組み込み, 十分な部品化が可能であることを示す。, 1997年09月24日, 55, 1, 331, 332
査読無し, その他, 情報処理学会研究報告. 計算機アーキテクチャ研究会報告, 社団法人情報処理学会, データ分割配置を考慮するループディストリビューション, 中西恒夫; 城和貴; PolychronopoulosConstantine D; 荒木啓二郎; 福田晃, 本稿では, 自動並列化コンパイラの中間表現であるデータ分割グラフ上で, データ分割配置とループディストリビューションを同時に行うアルゴリズムを提案する. 従来の一般的なループディストリビューションアルゴリズムでは, ループネストは可能な限り細かく分割され, また並列化なループは全て並列化される. しかしながら, 通信オーバーヘッドが比較的大きな分散メモリ型並列計算機上では, 過剰なループネストの分割と並列化は, プロセッサ間通信のために当該ループネストの実行時間が引き延ばされる結果に陥る. 提案アルゴリズムでは, データ転送コストを考慮の上で適正なループネストの分割と並列化を行う., 1997年08月20日, 97, 76, 13, 18
査読無し, 日本語, 情報処理学会研究報告. [ハイパフォーマンスコンピューティング], 社団法人情報処理学会, オブジェクト指向の枠組によるループ並列化記述, 大森洋一; 城和貴; 福田晃, 並列化コンパイラは従来のコンパイラより複雑であるため, オブジェクト指向設計に基づく改善を試み, 特にループ並列化作業の対象となるループ中間表現に必要な情報を抽出し, クラス化する. この際, ループ並列化全体の枠組として, 作業を3つのフェーズに分割し, これらの繰り返しによって最適化を進めて行く設計法を提案し, ハードウェア情報の取り込みを行なうようにする. これにより, アルゴリズム記述が基本的なデータ構造と分離され, 拡張性・保守性が向上するとともに, ハードウェアの条件を利用して, 定量的な最適化を行なうことが可能になる., 1997年08月19日, 97, 75, 115, 120
査読無し, 日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 並列化支援視覚化システム NaraViewにおけるビュー間の連携方式について, 笹倉万里子; 木和田智子; 城和貴; 荒木啓二郎, 自動並列化コンパイラは、逐次処理用に書かれたプログラムを並列実行用に自動的に書き換えることを目的としている。しかし、対並列実行に関係するさまざまな条件をすべて考慮して最適な並列化を行なえるような評価方法は知られていない。そこでわれわれは、ユーザに、コンパイラがプログラムを解析して得た情報を視覚化して提示し、ユーザの指示によって並列化を行なうような並列化支援視覚化システムNaraViewを提案する。本稿では、特にNaraViewにおけるビューと、そのビュー間の連携について述べる。, 1997年01月24日, 97, 11, 9, 16
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, Communication-Parallelism Graphによるデータ自動分割手法, 齊藤哲哉; 城和貴; 福田晃, 逐次のアプリケーションをワークステーションクラスタのような分散環境の上で分散化しようとするとき、データをどのように分割し配置するかが大きな問題となる。分散化を行うソフトウェアの多くは、これらの作業を利用者に任せているが、自動的に行う手法もいくつか提案されている。本稿では、データの分割と配置を一度に決定することができる通信-並列性グラフ(communication-Parallelism Graph)を用いて、逐次のアプリケーションをリモートプロシージャコールベースの分散化コンパイラSCIDDLEでコンパイルできる形式に変換するトランスレータの実装方針について述べる。, 1997年01月24日, 97, 11, 25, 30
査読無し, 日本語, 電子情報通信学会論文誌, ループ最小並列実行時間を算出する一手法, 城和貴; 中西恒夫; Polychronopolos, C; 福田晃; 荒木啓二郎, 1997年, J80-D-I, 9, 763-773
査読無し, その他, 知能情報・高度情報処理シンポジウム報告書, 1997, cc-COMAの概要, 城和貴, 1997年
査読無し, その他, 並列処理シンポジウムJSPP'97 (チュートリアル講演), Sep, 情報処理学会, メモリ・コンシステンシ・モデルの諸定義と解釈例, 城和貴, 1997年
査読無し, その他, 並列処理シンポジウムJSPP'97(チュートリアル講演), 情報処理学会, メモリ・コンシステンシ・モデルの諸定義と解釈例, 城和貴, 1997年, 149, 163
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 分散型構造解析手法の負荷分散への適用, 小板隆浩; 城和貴; 福田晃, 分散システムを有効に利用するためには, システム内の各ノード間の負荷の違いを均等にする負荷分散が必要となる. 分散管理型の負荷分散は, それぞれのノードにおいて負荷を管理し, 負荷の転送を行ない, 大規模システムへの適用において有効である. 我々は、大規模分散システムの負荷分散手法として、構造物の変形解析として定式化されている分散型の解析手法を、負荷分散の問題に応用し、問題を定式化する。この構造物の解析手法は、隣接するノード間の部分通信により実現されており、大規模システムの負荷分散手法として応用も可能である. 提案手法をシミュレーシヨンにより評価した., 1996年11月26日, 96, 118, 41, 47
査読無し, 日本語, 情報処理学会研究報告. [ハイパフォーマンスコンピューティング], 社団法人情報処理学会, 並列化支援のためのデータ依存の3次元視覚化, 木和田智子; 笹倉万里子; 中西恒夫; 城和貴; 福田晃; 荒木啓二郎, 近年、グラフィックワークステーションの普及により、3次元を使った情報視覚化が多く行なわれるようになってきている。しかし、単に量的に多くの情報を表示するためだけでなく、3次元を使わなければ表示が難しい情報を視覚化した例はまだ少ない。本稿で、我々は、3次元を有効に使った視覚化法の一つとして、並列化支援のためのデータ依存関係の視覚化法を提案する。この方法は、3次元空間内に複数の変数のデータ依存関係を一度に表示することができる。この方法で視覚化された図を見ることで、ユーザは、並列化を妨げるデータ依存や無駄に使われている変数/配列要素の発見を容易に行なえるようになる。, 1996年10月04日, 96, 97, 37, 42
査読無し, 日本語, 情報処理, 社団法人情報処理学会, 並列計算機の解析モデル, 城和貴; 福田晃, 1996年06月15日, 37, 6, 528, 535
査読無し, その他, 情報処理学会研究報告. 計算機アーキテクチャ研究会報告, 社団法人情報処理学会, ループによって運ばれる依存を有するループの並列実行時間の見積り, 中西恒夫; 城和貴; PolychronopoulostConstantine D; 福田晃; 荒木啓二郎, ループは並列性の重要な源であり、またプログラムの実行時間の大部を占める。そのためこれまで、マルチプロセッサ用に様々なループのイタレーション単位(中粒度)での最適化技法が提案、使用されてきた。一方、スーパースカラ技術をはじめとする、近年のプロセッサ内部における細粒度並列処理技術の普及により、ループボディ内での最適化技法が重要になりつつある。ループ全体の並列実行時間の下限は、こうした異なる粒度での最適化手法の効果を、統一的かつ定量的に評価するのに有用な尺度であろう。ループ全体の並列実行時間の下限は、ループを完全に展開したコードの依存グラフのクリテイカルバスのコストを求めることにより得ることができる。しかしながら、ループを完全に展開することは実用上極めて問題である。本稿では、問題を整数計画問題に帰着させ、このクリティカルパスのコストを、ループを一切展開することなく得る手法を提案する。さらに提案手法を実装、実際のプログラムに適用し、提案手法が実用的な時間でクリテイカルバスのコストを与えることを示す。, 1996年03月05日, 96, 23, 25, 30
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, SMCIモデルのアップデート型キャッシュ・コヒーレンス・プロトコルへの適用, 城和貴; 福田晃, 本稿では、既に我々の提案している、インバリデーション型キャッシュ・コヒーレンス・プロトコルを採用した並列計算機に対する解析モデル、Semi-markov Memory and Cache coherence Interference (SMCI)モデルの、アップデート型キャッシュ・コヒーレンス・プロトコルへの適用について報告する。本稿で提案するSMCI/Dragonモデルは、元のSMCIモデル同様、代表的なアップデート型プロトコルであるDragonを採用した並列計算機の性能予測を行なう。確率過程等を利用したこれまでの解析モデルでは、並列システムの大規模化に伴い、状態数の爆発的増化という避けがたい問題があった。SMCIモデルでは、状態数はシステムの規模に無関係に定まる。本稿で報告するDragonプロトコルに対する状態数は20である。, 1996年03月05日, 96, 24, 3, 8
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 整数計画問題としてのループ並列実行時間の下限算出問題, 中西恒夫; 城和貴; PolychronopoulosConstantine D; 福田晃; 荒木啓二郎, 並列処理の分野において、プログラムの並列実行時間の見積りは、静的スケジューリング、自動並列化等の応用で重要な問題である。特に、プログラム中のループには多くの並列実行可能部分が潜在するため、ループ部の並列実行時間の見積りに対する需要は大きい。ループ全体の並列実行時間の下限は、ループを展開したコードの依存グラフを生成し、そのクリティカルパスのコストを算出することにより得ることができるが、この方法はループの大きさに対するスケーラビリティの点で劣る。本稿では、ループを展開することなく、ループを展開したコードの依存グラフのクリティカルパスのコスト-すなわちループ全体の並列実行時間の下限-を算出する問題を、整数計画問題に帰貴させ、単体法ならびに分技限定法を用いて解く。これにより、問題解決にかかる時間はループの大きさに依存せず、大規模なループの並列実行時間の下限の算出が可能となる。, 1996年01月22日, 96, 10, 13, 18
査読無し, 日本語, 情報処理学会研究報告. [ハイパフォーマンスコンピューティング], 社団法人情報処理学会, HTGの最適化手法への適応に関する考察, 丸川一志; 城和貴; CraigDavid; PychronopoulosConstantine; 福田晃, 我々の目標は、並列計算に関する研究のための環境としての、また基盤としての自動並列化コンパイラを構築することである。こめコンパイラは、最適化バスの独立性と、それらを自由に適用できる実装という特徴を持つ。そのようなコンパイラを構築するためには、中間表現をベースとして構築することが重要で、特に、中間表現とその利用手続きとを統一することが必要となる。本稿では統一された中間表現としてイリノイ大学CSRDで提案されたHTGを用いることについて検討を行う。また、最適化パスの実装を絡めて、利用手続きについての検討も行う。, 1995年10月18日, 95, 97, 1, 6
査読無し, その他, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 社団法人情報処理学会, 共有メモリ型並列計算機の解析モデル, 城和貴; 福田晃, キャッシュ・メモリを有した共有メモリ型並列計算機に対する、セミマルコフ過程を利用した解析モデルを提案する。提案されるSMCI(Semi-markov Memory and Cache coherence Interference)モデルは、キャッシュ・コヒーレンス制御命令や通常のデータ・リクエストの、ネットワーク競合やメモリ競合による待ち状態を、実際の並列プログラムに特有のパラメータを与えることにより、容易に記述できる。また、並列計算機自体のシステム構成も人力パラメータとして与えることができるため、さまざまなシステム構成を持つ並列計算機の、アプリケーションごとの性能評価を容易かつ詳細に得ることができる。これまでのマルコフ連鎖等を利用した並列計算機の解析モデルと異なり、SMCIモデルにおける状態数は、プロセッサ台数やメモリ・サービス時間に依存せず、評価対象とする並列計算機の採用しているキャッシュ・コヒーレンス・プロトコルによってのみ決定する。従って、SMCIモデルの計算時間は本質的にシステムの規模に左右されない。, 1995年09月20日, 95, 93, 13, 18
査読無し, 日本語, 情報処理学会研究報告. PRO, [プログラミング], 社団法人情報処理学会, 複合グラフを用いた階層タスクグラフの視覚化, 笹倉万里子; 木和田智子; 城和貴; 荒木啓二郎, 逐次プログラムを並列プログラムに書き換える自動並列化コンパイラの分野では,ユーザのプログラムの理解支援,デバッグ,並列性抽出の半自動化のためのインタフェースなどの目的のために並列化支援システムが必要とされてきている.本稿では,イリノイ大で研究開発されている自動並列化コンパイラParafrase-2の中間表現として使われているHTG(Hierarchical Task Graph)を軸としてソースレベルでのプログラムの依存情報を視覚化することで並列化支援を行なうシステムを提案する.特にHTGをグラフとして表示するための方法について述べる., 1995年08月24日, 95, 82, 25, 32
査読無し, その他, 情報処理学会研究報告. PRO, [プログラミング], 社団法人情報処理学会, OMT法による並列化コンパイラ中間言語フレームワークの構築, 大森洋一; 城和貴; 福田晃; 荒木啓二郎, 現在、コンパイル時におけるプログラム並列化手法が数多く提案されているが、それらの間の関係は必ずしも明らかでない。このため、各手法の組合せあるいは相互評価ができにくい。そこで、今後の研究を進めるにあたり、共通の条件により比較・検討できる環境が望まれる。この環境は、1.実装技術に左右されにくい。2.理論的に明解である。3.汎用性が高く、安定している。といった性質を備えているのが望ましい。本稿では、オブジェクト指向設計法の1種であるOMTを用いて、それぞれの手法を系統的に整理し、中間言語のフレームワーク生成の可能性を検討する。特に、コンパイラの中心となる意味解析木について解析結果を述べる。, 1995年08月24日, 95, 82, 33, 40
査読無し, その他, 情報処理学会研究報告. マルチメディア通信と分散処理研究会報告, 社団法人情報処理学会, ワークステーションクラスタにおけるSCIDDLEライブラリの評価, 齊藤哲哉; 城和貴; Luthi PeterHans; ArbenzPeter; 福田晃; 荒木啓二郎, SCIDDLEは、並列アプリケーションを並列コンピュータやワークステーションクラスタといった非均質なネットワーク環境に分散化するツールである。我々は、次世代のスーパーコンピューティング環境構築のための第一段階として、SCIDDLEの性能評価を行った。これにより、非同期呼び出しの制御を行う関数がシステムの性能に大きく影響することがわかった。また、実際にSCIDDLE上で動作する化学計算アプリケーションDISCOについても性能評価を行い、SCIDDLEがアプリケーションの性能にどのような影響を与えているかを調べた。これらの結果から、筆者らの提案する次世代のスーパーコンピューティング環境、すなわち、広域分散局所並列処理環境構築への指針を示す。, 1995年07月13日, 95, 61, 91, 96
査読無し, 日本語, 情報処理学会研究報告. [ハイパフォーマンスコンピューティング], 社団法人情報処理学会, HDPG:階層データ分割グラフ, 中西恒夫; 城和貴; PolychronopoulosConstantine D; 福田晃; 荒木啓二郎, 本稿では,並列化コンパイラの中間表現として階層データ分割グラフHDPGを提案する.従来のデータ分割手法は一般的にループ並列性に応じたものが多く,その分割の最適化は特定のループについて局所的になされる.そのため,複数のループを考えた場合は個々のループでの分割の間に不整合が生じがちである.HDPGは,我々がすでに提案しているデータ分割グラフDPGをプログラムの制御構造に応じて階層化したもので,異なる粒度における機能並列性に応じた,大域的なデータ分割手法を提供する土台としての機能する.HDPG上では,メモリ階層等のハードウェアの階層性と親和性の高いデータ分割が可能である., 1994年07月22日, 94, 68, 89, 94
査読無し, 日本語, 情報処理学会研究報告. [プログラミング-言語基礎実践-], 社団法人情報処理学会, 線形依存ベクトルのループの並列性抽出法, 北須賀輝明; 城和貴; 福田晃; 荒木啓二郎, 本稿では,剰余類による分割を用いた,ループ並列化手法と通信を考慮したイタレーションのプロセッサ割当手法を提案する.本手法の対象とするループは依存ベクトルが定数ではなく線形となる場合である.既存の手法では依存ベクトルを定数に変換することで並列化されている.そのため,並列性が犠性にされ,プロセッサ間通信に対する考慮がなされていない.本手法では,剰余類を用いてループを分割し,これを単位として並列化とプロセッサ割当を統一的に行う.剰余類による分割により,イタレーション数nに対してO(log(n))の並列化および通信を意識したプロセッサ割当が可能となる., 1994年07月21日, 94, 65, 25, 32
査読無し, その他, 情報処理学会研究報告. 計算機アーキテクチャ研究会報告, 社団法人情報処理学会, DPG:データ分割グラフ, 中西恒夫; 城和貴; 福田晃; 荒木啓二郎, 本稿では,並列化コンパイラの中間表現としてデータ分割グラフDPGを提案する.これまで,中間表現として用いられてきたタスクグラフは,タスク間依存関係は表現されていたが,データ依存の源である変数そのものは表現されておらず,データ分割には適用し難いものであった.一方,データ分割はad-hocな手法が多く,また本来同時に考慮されるべき,スケジューリングなどのプロセッサ間の通信コストに絡む問題とは分けて論じられる嫌いがあった.DPGは変数をグラフ中に表現することにより,一般性の高いデータ分割手法を提供し,さらにはデータ分割をプロセッサ間通信コストの種々の問題と同時に取り扱うための土台を提供する., 1994年01月27日, 94, 13, 121, 128
査読無し, 日本語, 情報処理学会研究報告. [ハイパフォーマンスコンピューティング], 社団法人情報処理学会, 並列計算機の解析モデル : 共有ブロックの予測, 城和貴; 福田晃, キャッシュを有した共有メモリ型並列計算機の解析モデルを構築する際に,キャッシュ・コヒーレンス制御のモデル化は重要かつ構築困難なところである.これまでに提案されているモデルには,データを共有ブロックとプライベート・ブロックに分離することにより,共有ブロック,すなわちコヒーレンス制御のモデル化を試みているものがある.このモデルは簡潔である反面,プライベート・ブロックへのミス・ヒットが共有ブロックに影響を与えない,という前提条件が必要であった.しかしながら,並列計算機における実際のアプリケーションを考えた場合,プライベート・ブロックへのミス・ヒットが無視できない規模のプログラムも多い.そこで本稿では,プライベート・ブロックへのミス・ヒットが共有ブロックに影響を与えることを考慮した,並列計算機の解析モデルを提案する.マルコフ連鎖を用いて構築された本解析モデルでは,プライベート・ブロックに対するキャッシュ・ヒット率を与えることで,共有ブロックがどのような状態に分布するかを予測する., 1993年08月19日, 93, 72, 81, 88
査読無し, 日本語, 情報処理学会研究報告. [ハイパフォーマンスコンピューティング], 社団法人情報処理学会, アドレストレースを利用した並列計算機のパラメトリックシミュレータ, 大森洋一; 城和貴; 福田晃; 荒木啓二郎, 従来,既存のマルチプロセッサ以上の規模のマシンの評価は,解析モデルもしくは確率によるシミュレーションで行なわれることが多かった.しかし,実マシンの動作にはこれらでは表現しきれない部分がある.これに対し,トレースドリブンなシミュレータは実際を忠実に表現できるが,実行時間・必要なデータ量から大規模マルチプロセッサへの適用が困難である.そこで,トレースドリブンを基本に,他のプロセッサと独立な部分は確率モデルを適用する.これにより,従来のトレースドリブン型のシミュレーションに比べ,大幅にメモリ使用量を減らすことができた.また,データ操作の省略により,シミュレーション時間も短縮された.これによる動作の相違は5%程度であった., 1993年08月19日, 93, 72, 113, 120
査読無し, 日本語, 情報処理学会研究報告. 計算機アーキテクチャ研究会報告, 社団法人情報処理学会, 並列計算機の解析モデル : シミュレーションとの比較, 城和貴; 福田晃, キャッシュを有した共有メモリ型並列計算機に対する,セミ・マルコフ過程を利用した解析モデルを提案する.提案されるモデルはコヒーレンス制御やネットワーク競合による待ち状態を,実際の並列プログラムに特有のパラメータを与えることにより,容易に記述できる.また,並列計算機自体のシステム構成も簡単に変更できるため,さまざまなシステム構成を持つ並列計算機の,アプリケーションごとの性能評価を容易かつ詳細に得ることができる.さらに,構築されたモデルを用いて,プロセッサ利用率,通常データ/コヒーレンス制御リクエスト及びそれらの待ち時間について実際の評価を行なう.その結果,実際のシミュレーション結果と較べて4.36%の違いしかないにもかかわらず,計算時間は普通のワークステーションでわずか0.01秒程度であった., 1993年06月11日, 93, 49, 17, 24
査読無し, その他, 生物の科学「遺伝」別冊, 裳華房, ノイマン型からニュ-ロコンピュ-タ-へ (脳研究とニュ-ロコンピュ-タ--学習・記憶のメカニズムを探る) -- (ニュ-ロコンピュ-タ), 三宅誠; 城和貴, 1989年11月, p96, 101
査読無し, その他, 89'信学秋全大,SD-11-4, 階層型ニューラルネットの統合, 城和貴, 1989年, 6.282, 6.283
査読無し, その他, テレビジョン学会誌, 社団法人映像情報メディア学会, 5)読唇による単音節の認識([画像表示研究会視覚情報研究会画像通信システム研究会]合同), 積山薫; 城和貴; 梅田三千雄, 1988年03月20日, 42, 3
査読無し, その他, 信学技法,PRU88-94, ニューラルネットの学習における新しいタイプの教師, 城和貴, 1988年
査読無し, その他, 信学技報, PRU88-94, ニューラルネットワークの学習における新しいタイプの教師, 城和貴, 1988年
査読無し, 日本語, テレビジョン学会技術報告, 社団法人映像情報メディア学会, 読唇による単音節の認識 : 画像表示 : 視覚情報 : 画像通信システム(<特集>視覚と画質), 積山薫; 城和貴; 梅田三千雄, Human lipreading abilities were investigated. A VCR (VTR) was used to display a female narrator who spoke 100 Japanese syllables. The subjects were 40 normal adults. They were assigned to one of two narrators and participated in 3 experimental sessions of 100 syllables. The result showed: (a) Although only 15% of the syllables were reported correctly, 80% of vowels and 20% of consonants were read correctly. (b) By overlooking some types of errors, such as confusion within labials, within voiced and voiceless pairs, and so forth, 35% of syllables were recognized correctly. (c) Labials were a..., 1988年01月, 12, 1, 33, 40, 10.11485/tvtr.12.1_33
査読無し, その他, 情報処理学会論文誌, 情報処理学会, 「情報処理学会論文誌:数理モデル化とその応用」の編集にあたって, 城 和貴; JOE Kazuki; 北 栄輔; KITA Eisuke, 48
その他, 日本ソフトウエア科学会大会論文集, タンパク質分子構造可視化システムProteinViewの試作, 平尾智也; 水原隆道; 斉藤哲哉; 安藤誠; 秋山泰; 城和貴; 国枝義敏, 1999年, 16th
日本語, 研究報告ハイパフォーマンスコンピューティング(HPC), 疎行列のキャッシュ適合性に基づくGraph500ベンチマークの特性解析, 田邊 昇; 冨森 苑子; 高田 雅美; 城 和貴, Graph500 はビッグデータ解析や疎行列処理のベンチマークとも言われており,近年注目を集めている.本論文では Graph500 の疎行列や,その Randomize や先行研究の Vertex sorting の有無を制御した行列のキャッシュメモリへの適合性を,空間的局所性の観点から解析した.その結果,Graph500 の疎行列の空間的局所性は極めて低く,疎行列ベクトル積 (SpMV) を行う場合にはベクトルのアクセスの際に,キャッシュラインには平均して 1 個程度しか有効なデータは載っていないことが判った.ランダマイズによりその傾向が強まる.Vertex sorting は空間的局所性にはあまり作用しなかった.上記行列の SpMV 実行時の GPU の L1 キャッシュヒット率は,Vertex sorting の効果が小さく,グラフサイズ (SCALE) が大きくなるほど低下した.データサイズの大きさと非零要素配置のランダム性のため,キャッシュアウェアなソフトウェア最適化が困難である.特に空間的局所性の低さに伴う問題は現時点では未開拓な性能改善要因である.ゆえに,Graph500 や Green Graph500 の競争で優位に立つには,1 桁程度のメモリトラフィック削減による高速化や省電力化が期待できるメモリ側の Gather 機能が有望である.Graph500 is a benchmark suite for big data analysis and sparse matrix processing which receives attention in these years. The spatial locality of sparse matrices used for Graph500 and their variations which are controlled in adoption of randomization and vertex sorting which is a technique of preceding work are investigated. We show the spatial locality of sparse matrices used for Graph500 is very low and there is about 1 or a little more valid data on a cache line for the memory accesses issued by sparse matrix-vector multiplications (SpMV) in average. The tendency is increased by randomization. The effect of vertex sorting does not have big effects on spatial locality. Effect of vertex sorting on L1 hit rate of GPU on processing SpMV for the above matrices is small and is degraded when the size of graph (i.e. SCALE) becomes larger. It is very difficult to solve the problem by just software approach because of the huge size of sparse matrices and the randomness of their accesses to degrade the optimization against their cache awareness. Especially problems caused by low spatial locality are frontiers of improvement, now. Therefore, hardwired gather functions at memory side is promising for taking advantage in the Graph500 lists, which improve the processing speed due to reduction of memory traffic in an order of magnitude., 2013年02月14日, 2013, 30, 1, 7
日本語, 情報処理学会研究報告. MPS, 数理モデル化と問題解決研究報告, 一般社団法人情報処理学会, ニューラルネットワークを用いた心拍データの学習と利用, 赤尾 維; 野村 涼子; 石川 由羽; 高田 雅美; 城 和貴, 近年,少子高齢化の急速な進展や生活習慣病の増加に伴う医療負担の増大が問題となっており,個人で健康管理を行う予防医療が求められている.そのためにリアルタイムで健康を管理できるウェアラブル機器が開発されている.また心拍変動の解析によって心理的ストレスの推定ができることが知られている.心理的ストレスの推定には安定した心拍データが求められるため,ノイズの除去や欠損値の補完が重要である.日常生活においてウェアラブル端末で心拍を取得する際,ノイズの混入や一時的にデータを取得出来ない場合がある.そこで本研究ではニューラルネットワークを用いて心拍の学習を行い,心拍データの欠損を個人の個性を残して補完する手法を提案する., 2015年02月24日, 2015, 17, 1, 6
日本語, 研究報告コンシューマ・デバイス&システム(CDS), 一般社団法人情報処理学会, 定在波レーダを用いたリアルタイム呼吸モニタリングシステムの開発, 高木 里穂; 石川 由羽; 高田 雅美; 梅田 智弘; 城 和貴, 乳幼児突然死症候群は,1 歳までの健康な乳児が予兆なしに突然呼吸停止し,死亡する症状のことである.この症候群を予防するには呼吸循環系の異常を早期に発見する必要がある.そこで,呼吸循環系の異常を検出するために定在波レーダを用いる.定在波レーダは物体との変位を計測することができる.この定在波レーダを用いて,呼吸による体表面の振幅の変位を取得する.取得した変位に周波数解析を行い,乳児の呼吸の有無をリアルタイムに観察する呼吸モニタリングシステムの開発を行う.Sudden infant death syndrome is a disease that a well-looking baby of up to one year stops his/her breathing without any reason, and comes to die. To prevent babies from this disease, some kind of systems that check baby's breathing circulating system to detect abnormal foretaste is required. Such error detection for baby's breathing circulating can be realized by standing wave radars, which measures the distance of target objects precisely. Measuring series of displacements to the breast of a baby by using standing wave radars. we apply Fourier transform analysis for the frequency band of baby's breathing, and we present the development of respiration monitoring system to observe of baby's breathing., 2015年01月19日, 2015, 1, 1, 6
日本語, 研究報告グループウェアとネットワークサービス(GN), 一般社団法人情報処理学会, 定在波レーダを用いたリアルタイム呼吸モニタリングシステムの開発, 高木 里穂; 石川 由羽; 高田 雅美; 梅田 智弘; 城 和貴, 乳幼児突然死症候群は,1 歳までの健康な乳児が予兆なしに突然呼吸停止し,死亡する症状のことである.この症候群を予防するには呼吸循環系の異常を早期に発見する必要がある.そこで,呼吸循環系の異常を検出するために定在波レーダを用いる.定在波レーダは物体との変位を計測することができる.この定在波レーダを用いて,呼吸による体表面の振幅の変位を取得する.取得した変位に周波数解析を行い,乳児の呼吸の有無をリアルタイムに観察する呼吸モニタリングシステムの開発を行う.Sudden infant death syndrome is a disease that a well-looking baby of up to one year stops his/her breathing without any reason, and comes to die. To prevent babies from this disease, some kind of systems that check baby's breathing circulating system to detect abnormal foretaste is required. Such error detection for baby's breathing circulating can be realized by standing wave radars, which measures the distance of target objects precisely. Measuring series of displacements to the breast of a baby by using standing wave radars. we apply Fourier transform analysis for the frequency band of baby's breathing, and we present the development of respiration monitoring system to observe of baby's breathing., 2015年01月19日, 2015, 1, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, 機械学習を用いた問題解答のための推論システムの開発, 岩片 悠里; 川嵜 美波; 江口 由記; 石川 由羽; 高田 雅美; 城 和貴, 本稿では,問題文の解答を推測するシステムの開発について述べる.本システムでは,問題文の言語解析及び特徴ベクトルの抽出によって学習データの作成を行った後,機械学習を行い問題文のパターン分類を行うことによってテストデータから問題文の答えを推測することができる.機械学習には,教師あり学習を用いるべきである.多クラス判別を行う教師あり学習の代表的な手法に Learning Vector Quantization がある.Learning Vector Quantization では事前にクラス数を求めておく必要がある.クラス数を求める方法として,カーネル主成分分析及びギャップ統計量がある.ただし,カーネル主成分分析は視覚的に推定する方法であり,ギャップ統計量は,教師なし学習のための推定方法であるため,本システムには適さない.一方,Affinity Propagation はクラス数を事前に求める必要がないが,教師なし学習の手法である.そこで,本稿では,Affinity Propagation および Learning Vector Quantization を複合させ,クラス数の推定を行う教師あり学習の手法を提案する.In this paper, we develop a system to guess the answer of the questions. In the system, once training data is created with language analysis and feature vectors of question statements, then a learning machine is performed for pattern classification. By using the system, the answer can be obtained. LVQ(Learning Vector Quantization) is known as a representative method of supervised learning for multi class classification. LVQ needs the number of classes as initialization. Kernel principal component analysis and gap statistic are methods for getting number of classes. Since kernel principal component analysis is a method for estimating visually and gap statistic is a method for unsupervised learning, there is not suitable for the proposed system. AP(Affinity Propagation),which is a method of unsupervised learning, is not necessary to determine the number of classes in advance. Hence, we propose a method for performed supervised learning to estimate number of classes through AP is improved using LVQ., 2014年12月02日, 2014, 10, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, ボリュームデータにおける三次元物体の部分認識, 長尾 英里; 石川 由羽; 高田 雅美; 城 和貴, 本稿では,ボリュームデータにおける 3 次元物体の部分認識の手法を提案する.病院の診察で利用される検査方法として,MRI(magnetic resonance imaging) や CT (Computed Tomography) で身体の内部を撮影する方法がある.これらはスライス画像群として撮影されている.このスライス画像を 3 次元構築し,可視化することにより,3 次元物体をイメージできる.また,抽出したい部位の選択を自動化することによって,病原の情報を素早く容易に診断に役立てることができるようになる.そのために,提案手法では,スライス画像群からボリュームデータを構築し,構築したボリュームデータから 3 次元 SIFT 特徴量を抽出する.抽出した特徴量を基に LVQ (Learning Vector Quantization) で識別する.In this paper, we propose a method of partial recognition for three dimensional object in volume data. In hospitals, MRI (magnetic resonance imaging) and CT (Computed Tomography), which take photos as slice images, are often adapted to obtain information in inner of body. By constructing slice images to a three-dimensional object, users understand information of slice images intuitively. By an automatic partial recognition, location of pathology can be clear speedy. In the proposed method, slice images are transformed into volume data and a partial recognition is performed using LVQ (Learning Vector Quantization) with the three-dimentional SIFT features., 2014年12月02日, 2014, 9, 1, 6
日本語, 電子情報通信学会技術研究報告 = IEICE technical report : 信学技報, 一般社団法人電子情報通信学会, 活字データの分類を用いた進化計算による近代書籍からのルビ除去 (情報論的学習理論と機械学習), 粟津 妙華; 高田 雅美; 城 和貴, 国立国会図書館では,所蔵する明治から昭和前期の近代書籍を近代デジタルライブラリとしてWeb上でページごとの画像データとして公開しているが,文書内容での検索を行うことができない.そのため,自動でのテキストデータ化が望まれている.その際,問題となっているのがヒストグラムでは除去できないルビであり,我々はすでに近代書籍に特化したルビ除去手法を提案している.しかしながら,提案した手法は書籍に付加された外部情報を元にしており,実現可能性は低い.そこで本論文では,書籍画像から直接得られるデータを元に,進化計算によってルビ除去式を生成し,近代書籍から自動でルビを除去する手法を提案する., 2014年06月25日, 114, 105, 117, 122
日本語, 情報処理学会研究報告. BIO, バイオ情報学, 一般社団法人情報処理学会, 活字データの分類を用いた進化計算による近代書籍からのルビ除去, 粟津妙華; 高田雅美; 城和貴, 国立国会図書館では,所蔵する明治から昭和前期の近代書籍を近代デジタルライブラリとして Web 上でページごとの画像データとして公開しているが,文書内容での検索を行うことができない.そのため,自動でのテキストデータ化が望まれている.その際,問題となっているのがヒストグラムでは除去できないルビであり,我々はすでに近代書籍に特化したルビ除去手法を提案している.しかしながら,提案した手法は書籍に付加された外部情報を元にしており,実現可能性は低い.そこで本論文では,書籍画像から直接得られるデータを元に,進化計算によってルビ除去式を生成し,近代書籍から自動でルビを除去する手法を提案する., 2014年06月18日, 2014, 20, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, 活字データの分類を用いた進化計算による近代書籍からのルビ除去, 粟津妙華; 高田雅美; 城和貴, 国立国会図書館では,所蔵する明治から昭和前期の近代書籍を近代デジタルライブラリとして Web 上でページごとの画像データとして公開しているが,文書内容での検索を行うことができない.そのため,自動でのテキストデータ化が望まれている.その際,問題となっているのがヒストグラムでは除去できないルビであり,我々はすでに近代書籍に特化したルビ除去手法を提案している.しかしながら,提案した手法は書籍に付加された外部情報を元にしており,実現可能性は低い.そこで本論文では,書籍画像から直接得られるデータを元に,進化計算によってルビ除去式を生成し,近代書籍から自動でルビを除去する手法を提案する.In National Diet Library, books which are possessed in library as "the digital library from meiji era" are open to the public on Web. Since these are shown as image data and cannot search using document contents, an automatic text conversion is needed. There is a major obstacle to text conversion. It is ruby. Ruby can not be removed in the histogram method. Therefore, we have proposed a ruby removal method for early-modern Japanese printed books. However, since the proposed method is based on the external information added to the books, the feasibility is low. In this paper, we propose a method to remove the ruby automatically from early-modern Japanese printed books by generating ruby removal formula in Genetic Programming using the training data was based on the data of book image., 2014年06月18日, 2014, 20, 1, 6
日本語, 電子情報通信学会技術研究報告 = IEICE technical report : 信学技報, 一般社団法人電子情報通信学会, CSoC (Configurable System-on-Chip)の位置付けと発展性について (ディペンダブルコンピューティング 組込み技術とネットワークに関するワークショップETNET2014), 松本 尚; 城 和貴, LSIの集積度が上がり、SoC(System-on-Chip)が現実的になった当初には,製品ごとや企業ごとに差別化のためのASIC(Application Specific IC)開発がまだ盛んであった.このため,今後はSoCタイプのASIC開発がますます盛んになると予想して,日本の半導体産業ではASIC開発能力を高める努力を行った.しかし,高集積LSIの開発費の高騰などの要因により,生き残ったのは一般顧客向けのLSIであるASSP(Application Specific Standard Product)であった.ASSPでありかつSoCであるLSIが現在では隆盛を極めている.このASSP SoCのうちの多くの製品は外部ピン数に対応するよりもはるかに多くの機能をチップ上に集積して,製造後に利用者が使用する機能の選択や外部ピンの役割を設定可能である.この特徴を持つASSP SoCをCSoCと呼ぶことにする.半導体製品の中でCSoCが生き残ったことの必然性について論じ,CSoCの今後について議論する。, 2014年03月15日, 113, 498, 157, 162
日本語, 研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, 多フォント漢字認識手法における各カテゴリと必要教師データ数の分析, 粟津妙華; 福尾真実; 高田雅美; 城和貴, 近代書籍の漢字のフォントは現在のものとは大きく異なる特徴を持つため,既存 OCR を用いた認識率は低い.既に我々は近代書籍に適したオフライン多フォント漢字認識手法を提案している.本論文では,1000 種の漢字に多フォント漢字認識手法を適用し,漢字の特徴と必要教師データ数の関係性を分析する.我々の認識手法では PDC 特徴を用い SVM によって漢字を認識させるが,学習時において必要となる漢字と必要教師データ数の分析を十分に行っていなかった.そこで,PDC 特徴を用いて求めた漢字の特徴値と教師データ数による認識率の変化から,漢字の特徴値と必要教師データ数の関係性を分析する.Since the fonts of the Kanji character in early-modern Japanese printed books dating from the Meiji era to the first half of Showa era (AD1868-1945) has the greatly different feature from the standard fonts determined by JIS. We have been proposed the multi-fonts Kanji character recognition method suitable for early-modern Japanese printed books. In this paper, the multi-fonts Kanji character recognition method is applied to 1000 kinds of Kanji characters, and the relationship the between the feature of Kanji characters and the numbers of necessary training data is analyzed. In our proposed method, when Kanji characters are recognized with the PDC feature and SVM, the training data is needed. However, it was not analyzing about the relation between Kanji characters and the number of necessary training data. Therefore, the reration between the feature of Kanji characters and the number of necessary training data is analyzed using the feature value of Kanji characters by PDC vector and the change of recognition rate for each category of training data., 2014年02月24日, 2014, 10, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, 脳波個人認証のための特徴抽出, 石川 由羽; 吉田 智奈美; 高田 雅美; 城 和貴, 本稿では脳波を用いたバイオメトリック認証に有効な特徴抽出について検討する.個人の特徴量としてパワースペクトルを使用し,θ 波,α 波,β 波,γ 波の周波数帯域別に特徴量を抽出する.多チャンネル脳波計で 5 つのタスク時の脳波を計測し,それぞれの誤り率の解析を行う.結果,θ 波の誤り率が最も低く,イメージタスクでは右半球,後頭部の誤り率が低くなった.16 チャンネル 5 タスクでの誤り率は 3.8%であった.In this report we investigate feature extraction methods effective for biometric identification using brain waves. We extract power spectrum of theta, alpha, beta and gamma waves for the quantity of personal characteristic. We measure the brain waves of five tasks in multichannel electroen-cephalograph and analyze each error rate. As a result, the theta wave generates the lowest error rate, especially for the right hemisphere and occipital lowered in the image task case. The error rate by 16 channel and 5 tasks was 3.8%., 2014年02月24日, 2014, 20, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, オントロジーを用いた建造物のバリアフリー度判定, 松本 百世; 高田 雅美; 城 和貴; 城 和貴, 本研究では,オントロジー技術を用いてバリアフリー度の判定を行うシステムを構築する.システム開発にあたり,建造物のオントロジーと建造物内における人の行為のオントロジーをそれぞれ作成する.建造物オントロジーは建造物の構造に着目し,行為オントロジー行為主により生じる行為の差異に着目して作成する.開発するシステムには,オントロジーの情報に応じて,バリアフリー度評価に必要な指標を設ける.2 つのオントロジーを組み合わせて利用することで,様々な人視点でのバリアフリー度の判定を可能にする.In this research, we construct a system for barrier-free level determination using ontology. The system includes a building ontology and an action ontology. The buliding ontology targetes construction of a building. In the action ontology, actions to the buliding is differ each users characters. In the system, reference indeces for barrier-free level are set suitable based on these ontologies. By using two ontologies, the system can target anonimous users., 2014年02月24日, 2014, 15, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, 複数の3軸加速度センサを用いたNon-Exercise Activity Thermogenesisを含む身体活動のモデル化, 中尾 京子; 梅田 智広; 高田 雅美; 城 和貴, 日常生活における運動強度の低い身体活動 (NEAT) による代謝は,人の総エネルギー消費量の 20~30%を占め,個人差も大きい.そこで,運動だけでなく,NEAT を含むあらゆる身体活動のエネルギー消費量を推定するモデル化を行う.エネルギー消費量は体重に比例するため,体の中で最も比重の大きい胴体のエネルギー消費に注目する.胴体の動きをより正確に捉えるため,両胸椎,腰椎の 3 点の 3 軸加速度データを用いる.胴体の 3 軸加速度から得られた運動量,運動エネルギー,力積,仕事量から,NEAT を含む身体活動によるエネルギー消費のモデルの提案を行う.The metabolism by physical activity of low exercise intensity in daily life (Non-Exercise Activity Thermogenesis: NEAT), account for 20-30% of total energy expenditure of the people, its individual difference is large. Therefore, we propose the modeling for estimating energy expenditure of any physical activities including not only exercise but also NEAT. Energy expenditure is in proportion to body weight. Accordingly we model by limiting the energy expenditure of the body having a large specific gravity in the whole body. To capture more accurate movement of the body, we use three-axis acceleration data of the three points, right thoracic and left thoracic and lumbar. We propose a model of energy expenditure by physical activity including NEAT from momentum, kinetic energy, impulse, and work., 2014年02月24日, 2014, 21, 1, 6
日本語, 研究報告ハイパフォーマンスコンピューティング(HPC), 一般社団法人情報処理学会, 疎行列ベクトル積性能を決める諸要因, 田邊昇; 冨森苑子; 高田雅美; 城和貴, 疎行列ベクトル積 (SpMV) は多くの場合にキャッシュアーキテクチャとの相性が悪い.並列処理においては負荷不均衡が性能に与える影響も大きい.これまでは SpMV 性能を決める要因として,キャッシュのヒット率や一行あたりの非零要素数の平均,最大値,分散が注目されていた.しかし,それらと性能との相関が不明瞭であり,SpMV の挙動は長年にわたり謎に包まれていた.それは SpMV の最適化や,効率的な疎行列ライブラリ構築の障害であった.本報告では,SpMV 性能を左右する様々な要因をアプリケーション依存の要因とプラットフォーム依存の要因に分けて考察した.それを踏まえて行列の非零要素配置から導かれる時間的局所性と空間的局所性等のアプリ依存パラメータを導入した SpMV 性能モデルを構築した.その上でフロリダ大コレクションから抜粋した 115 種の疎行列と GPU を用いて SpMV 性能モデルの評価実験を行った.その結果,GPU 上で実行する場合は Padding に関する補正と小さな行列での補正が必要であることと,長行を折り畳むなど適切な負荷分散がなされた場合はキャッシュのヒット率よりも,空間的局所性やインデックス転送の抑制の方が実効性能に敏感であることが明らかになった., 2014年02月24日, 2014, 7, 1, 10
日本語, 研究報告コンピュータビジョンとイメージメディア(CVIM), 一般社団法人情報処理学会, 3次元形状データにおける色段差を利用した特徴抽出法の提案, 柳澤 希美; 高田 雅美; 城 和貴, 本論文では色段差による凹凸を形状特徴として抽出する手法を提案する.この凹凸は計測表面上において反射率が大きく異なるような境界で発生する.反射率は色の違いが原因であるため,色段差と呼ばれる.色段差は 2D 画像において輪郭エッジが検出される位置と同一である.色段差による凹凸と 2D 画像における輪郭エッジを位置合わせすることで,正確なテクスチャマッピングが可能となることが期待できる.3D スキャナ VIVID910 での計測により得られたポリゴンメッシュモデルに対し,本手法を適用し,色段差による凹凸を抽出することで本手法の有用性の確認を行う.In this paper, we propose a feature extraction method using color bound in order to find concavo-convex shape. The concavo-convex shape is constructed to border lines, of which reflectance rates are widely difference in a surface of a measurement 3D model. Since each color has a difference reflectance rate, the border lines are called as color bounds. A position of a color bound in a 3D model becomes equal to the edge position related to the color bound in a 2D picture. Therefore, through the position of the color bound is combined with the edge position, texture mapping can be performed exactly. To evaluate the proposed method, we adopt to 3D models, which are obtained by VIVID910., 2013年11月21日, 2013, 20, 1, 6
日本語, 研究報告ハイパフォーマンスコンピューティング(HPC), 一般社団法人情報処理学会, アーキテクチャの進歩を促進するGraph500ベンチマークのあり方, 田邊昇; 冨森苑子; 高田雅美; 城和貴, Graph500 はビッグデータ解析や疎行列処理のベンチマークとも言われており,Top500 を補うものとしても近年注目を集めてきた.従来の Graph500 参照実装は従来の計算機が苦手とする疎行列ベクトル積系応用全般を代表し,計算機の発展を促進する力があった.一方,ビットマップ式訪問管理と Hybrid BFS アルゴリズムがデファクトになる流れにある.筆者らの予備調査の結果,それらを実装した Pre release 版 Graph500 はキャッシュに適合し,メモリアクセス負荷が大幅に軽減され,本来とは異質なワークロードになることが明らかになってきた.本報告では,計算機の発展を促進する観点から,スーパーコンピュータのランキングに利用されるベンチマークの役割と,その望ましいあり方を考察する.さらに Hybrid 動作における局所性とキャッシュ挙動の観点から上記の予備調査結果を補強するとともに,Graph500 の将来について考察する., 2013年09月23日, 2013, 18, 1, 9
日本語, 研究報告ハイパフォーマンスコンピューティング(HPC), 一般社団法人情報処理学会, Graph500のHybrid解法に内在する局所性, 田邊 昇; 冨森 苑子; 高田 雅美; 城 和貴, Graph500 はビッグデータ解析や疎行列処理のベンチマークとも言われており,近年注目を集めている.ビットマップ式の訪問管理方式や,疎行列の一部だけアクセスする Beamer の Hybrid BFS アルゴリズムの公開により,今後はこれらを用いることがデファクトになる流れにある.本研究ではこれら 2 つを取り入れた最新の Graph500 のOpenMP 参照実装 Pre release 版のメモリアクセス特性を,局所性の観点から解析した.その結果,従来法よりもメモリアクセス回数,ミス時のダメージの大きさを示す列インデックス列の空間的局所性指標,ミスの起きにくさを示す時間的局所性 (平均ワーキングセットサイズ) が劇的に変化してしまっており,メモリアクセス回数削減率や時間的局所性がグラフサイズに比例して楽な状況になる傾向を確認した.つまり,従来版 Graph500 が従来の計算機では苦手な処理である疎行列ベクトル積系応用全般を代表し,計算機の発展を促進する意義のあるベンチマークであったのに対し,Hybrid 解法を入れた Graph500 は BFS の最適実装のみの特性を測るベンチマークに変質することが判った.さらに処理時間のホットスポットから BFS を行う関数が消え去り,実質的に検索の速度を測るベンチマークではなくなることが判った.ベンチマークの本来の意義を考慮して,それを損なわない適切な実施ルール作りを行なうべきことを提唱する.Graph500 is a benchmark suite for big data analysis and sparse matrix processing which receives attention in these years. Since the use of a bitmap to mark the visit of vertices and Beamer's Hybrid BFS algorithm to make a part of a sparse matrix be accessed are reported, we believe that such methods are to be a de facto standard. In this paper, from the viewpoint of reference locality, we analyze the memory access characteristics of pre-release version implemented with OpenMP and the two methods for the latest Graph500 reference code. As the results, we confirm that the number of memory accesses, a spatial locality metric of row index that shows performance degradation in the case of cache misses, and temporal locality (average working set sizes) that shows the resistance to cache misses, change drastically. We confirmed that the reduction rate of the number of memory accesses and temporal locality are accelerated in proportion to the target graph sizes. It means that the Graph500 with Hybrid method is another benchmark suite to measure the characteristics of optimized BFS implementation, while the conventional Graph500 represents most applications with sparse matrix vector multiplications that are not efficiently performed on conventional computer systems to require the improvement of computer systems. Furthermore, the hot spots against processing times do not include the functions to perform BFS, and we find it is no longer the benchmark for search speed practically. To investigate the essential meaning of the benchmarks, we propose appropriate and practical rules that do not damage it., 2013年07月24日, 2013, 27, 1, 8
日本語, 研究報告計算機アーキテクチャ(ARC), Scatter/Gatherをどこで行なうべきか?-低電力で高スループットなランダムアクセスのために-, 田邊昇; 冨森苑子; 高田雅美; 城和貴, HPC における反復解法のみならず、近年ではビッグデータ解析のニーズの高まりにより、大量データへのランダムアクセスの低電力化・高スループット化のニーズが高まっている。キャッシュベースの CPU や GPU において、キャッシュから溢れる配列に対して Scatter/Gather を行なうと、ライン内の空間的局所性の欠乏により消費電力とスループットの両面で深刻な問題が発生する。特に電力やメモリスループットが不足する将来の大規模計算基盤においては、この問題が年々深刻さを増す。本報告では、上記の問題の解決策として、Scatter/Gather 機能を Hybrid Memory Cube 内で行なうことを提案する。提案方式の電力やスループットに関するメリットについて、モデルを構築しつつ、Graph500 ベンチマーク課題行列に対する疎行列ベクトル積を実例に考察する。, 2013年03月19日, 2013, 4, 1, 7
日本語, 年会論文集, 日本教育情報学会, 10J05 教材開発(奥行き学習教材(特許第5226821号))の報告と今後の課題(タブレット端末と教材開発,一般研究), 吉井 直子; Peterson Ward; 高田 雅美; 吉住 秀; 城 和貴, 算数・数学で学習する幾何分野,特に空間図形(3次元物体)の理解を高めるための教材開発・教材研究を行っている.初めに奥行き知覚の実験を行い,得られた実験データからオクリュージョン(重なり)効果を検証した.そこでオクリュージョンを用いた「奥行き学習教材」を開発し,2013年3月に特許権を取得した.本稿では特許発明の過程を報告し,教材の活用方法を紹介する.また知的財産権について述べる., 2013年, 29, 406, 407
日本語, 研究報告計算機アーキテクチャ(ARC), 時間的局所性を考慮した疎行列のキャッシュ適合性, 冨森 苑子; 田邊 昇; 高田 雅美; 城 和貴, エクサスケールマシンは複雑なメモリシステムとなることが予想されている.同マシンへの適用を視野に入れた疎行列ライブラリの実現に向け,本報告では疎行列のキャッシュへの適合性分類において,疎行列の形状から得られる列インデックス列の空間的局所性と時間的局所性に関する指標を併用することを提案する.さらに,提案指標をフロリダ大学の疎行列コレクションを用いて評価した.その結果,空間的局所性だけで説明できなかった諸現象が説明可能になった. L1 キャッシュヒット率と時間的局所性の指標値との間に相関を確認できた.さらに,両指標を用いたメモリアクセス機構選択基準を与える方程式を導出した.In Japan, memory system of ExaFLOPS machines is expected more complex. In this paper, we propose a new characteristic of sparse matrices about spatial locality and temporal locality of row-index sequences in order to classify suitability for cache memory systems. Moreover, we evaluate proposal characteristic using University of Florida Sparse Matrix Collection. As a result, it became possible to explain phenomenons which could not be explained only by spatial locality. It is confirmed that there are the correlations between temporal locality and the general purpose cache (L1) hit rate. Moreover, we developed the equation for access method selection based on proposed characteristics., 2012年12月06日, 2012, 28, 1, 8
日本語, 研究報告数理モデル化と問題解決(MPS), α波構成要素の切出しと解析, 石川 由羽; 高田 雅美; 城 和貴, α 波は 1929 年に Hans Berger によって発見された脳波であり,現在,多くの研究者によって研究されている.近年では,主に α 波の振幅や位相に注目が集まっており,様々な研究結果が報告されている.しかし,今現在, α 波がどのように構成されているかを示す具体的なモデル化は行われていない.本論文では α 波のモデル化を目指して, α 波の位相分析を行う.前処理として,脳波から α 波データを抽出し,そのデータを α 波構成波に分割する. α 波構成波の特徴から,振幅と波形幅は時間遅延の相関があることが認められる.Alpha waves are electroencephalogram discovered by Hans Berger in 1929 and have been studied by many researchers. Recently the amplitude and the phase of alpha waves attract attention, and various studies are reported. So far the alpha waves are not concretely modeled to explain how they are constituted. In this paper, we analyze the phase of alpha waves for their future modeling. As pre-processing, we extract alpha wave data from EEGs and divide the data into alpha wave constituent waves. Analyzing the feature quantity of the alpha wave constituent waves, it turns out that the amplitude and the wavelength of alpha wave constituent waves are correlated possibly with some time delay., 2012年11月29日, 2012, 34, 1, 6
日本語, 研究報告バイオ情報学(BIO), 書籍オントロジーのための推論部の開発, 岩片悠里; 川嵜美波; 江口由記; 高田雅美; 城和貴, 本稿では,教科書を基に作られたオントロジーから情報を得るための推論部の開発を行う.推論エンジンは既存のものを用い,推論規則を新たに定義する.この推論部では,ユーザから与えられた問題文から,答えの導出を行う.これにより,ユーザの負担を軽減させることを可能にする.さらに,専門書の知識を基に推論の知識モデルとなる書籍オントロジーを作成することによって,効率的な情報の取得を目指す.In this paper, we develop inference part for an ontology that is based on the textbook. We use an existing inference engine and define new inference rules. By using this inference part, the answer can be derived from a problem statement. Thus, it is possible to reduce the burden on users. The book ontology is construted based on the textbook as knowledge model of the inference. Therefore, it is possible to retrieve information efficiently., 2012年11月29日, 2012, 32, 1, 6
日本語, 研究報告バイオ情報学(BIO), 遺伝的プログラミングを用いた近代書籍からのルビ除去, 粟津妙華; 高田雅美; 城和貴, 国立国会図書館では,所蔵する明治から昭和前期の近代書籍を近代デジタルライブラリとして WEB 上でページごとの画像データとして公開しているが,文書内容での検索を行うことができない.そのため,自動でのテキストデータ化が望まれている.その際,問題となっているのがルビである.現在のルビを直線的に除去する技術は,規格に沿った現在の書籍を対象としたものであるため,現在の書籍とは違う特性を持つ近代書籍には適用できない.そこで,本研究では,遺伝的プログラミングを用いて,曲線的に出版者・時代ごとの専用ルビ除去式の生成を行う.In National Diet Library, books which are possessed in library as "the digital library from meiji era" are open to the public on WEB. Since these are shown as image data and cannot search using document contents, an automatic text conversion is needed. However, ruby is a disturbing text conversion. Since existing techniques of linearly removing ruby had developed for books of the current standard, the techniques are inapplicable to early-modern Japanese books, which have a specific characteristic different from characters of current books. In this paper, we propose a method to remove ruby from early-modern Japanese books using Genetic Programming., 2012年11月29日, 2012, 17, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 書籍オントロジーのための推論部の開発, 岩片悠里; 川嵜美波; 江口由記; 高田雅美; 城和貴, 本稿では,教科書を基に作られたオントロジーから情報を得るための推論部の開発を行う.推論エンジンは既存のものを用い,推論規則を新たに定義する.この推論部では,ユーザから与えられた問題文から,答えの導出を行う.これにより,ユーザの負担を軽減させることを可能にする.さらに,専門書の知識を基に推論の知識モデルとなる書籍オントロジーを作成することによって,効率的な情報の取得を目指す.In this paper, we develop inference part for an ontology that is based on the textbook. We use an existing inference engine and define new inference rules. By using this inference part, the answer can be derived from a problem statement. Thus, it is possible to reduce the burden on users. The book ontology is construted based on the textbook as knowledge model of the inference. Therefore, it is possible to retrieve information efficiently., 2012年11月29日, 2012, 32, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), LF/HFを用いた時空間ストレス指標の提案, 内村麻里奈; 江口由記; 川嵜美波; 吉井直子; 梅田智広; 高田雅美; 城和貴, 人間にとって快適な環境とはその環境下でどれくらいの割合で人がストレスを感じないかという指標で与えられるべきである.この環境とは時空間に渡るものであり,同じ場所でも時間帯や季節によってその指標は変動すべきものである.そこで,本論文では,ある時空間にいる多数の人間のストレスレベルを集約分析することで時空間ストレス指標を得ることを提案する.また,既に報告した個人のストレスレベルを時空間でリアルタイム表示させるツールの機能を拡張することで,多数の人間のストレスレベルを表示させるツールを試作する.このツールを利用することで,直観的に時空間ストレス指標を確認し,ストレスが高い場所を回避することもできる.よって,時空間ストレス指標を用いることで,ストレスを避けることが可能となり,より快適な生活を送ることができる.Comfortable environment for humans should be evaluated by indicator which is how many people do not have stress there. This environment represents spatiotemporal space. The indicator should change depending on time and season not even the same place. In this paper, we suggest a spatiotemporal stress indicator which is calculated for aggregating and analyzing many people's stress levels in spatiotemporal space. In addition we expand the tool, which displays individual stress levels in real time spatiotemporal space, to display other people's spatiotemporal stress levels. Using this tool, it is possible to avoid going to the place which hava a high stress level by confirming spatiotemporal stress levels intuitively. Thus, it helps to spend our healthy life by using a spatiotemporal stress indicator., 2012年11月29日, 2012, 2, 1, 6
日本語, 研究報告バイオ情報学(BIO), α波構成要素の切出しと解析, 石川 由羽; 高田 雅美; 城 和貴, α 波は 1929 年に Hans Berger によって発見された脳波であり,現在,多くの研究者によって研究されている.近年では,主に α 波の振幅や位相に注目が集まっており,様々な研究結果が報告されている.しかし,今現在, α 波がどのように構成されているかを示す具体的なモデル化は行われていない.本論文では α 波のモデル化を目指して, α 波の位相分析を行う.前処理として,脳波から α 波データを抽出し,そのデータを α 波構成波に分割する. α 波構成波の特徴から,振幅と波形幅は時間遅延の相関があることが認められる.Alpha waves are electroencephalogram discovered by Hans Berger in 1929 and have been studied by many researchers. Recently the amplitude and the phase of alpha waves attract attention, and various studies are reported. So far the alpha waves are not concretely modeled to explain how they are constituted. In this paper, we analyze the phase of alpha waves for their future modeling. As pre-processing, we extract alpha wave data from EEGs and divide the data into alpha wave constituent waves. Analyzing the feature quantity of the alpha wave constituent waves, it turns out that the amplitude and the wavelength of alpha wave constituent waves are correlated possibly with some time delay., 2012年11月29日, 2012, 34, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 遺伝的プログラミングを用いた近代書籍からのルビ除去, 粟津妙華; 高田雅美; 城和貴, 国立国会図書館では,所蔵する明治から昭和前期の近代書籍を近代デジタルライブラリとして WEB 上でページごとの画像データとして公開しているが,文書内容での検索を行うことができない.そのため,自動でのテキストデータ化が望まれている.その際,問題となっているのがルビである.現在のルビを直線的に除去する技術は,規格に沿った現在の書籍を対象としたものであるため,現在の書籍とは違う特性を持つ近代書籍には適用できない.そこで,本研究では,遺伝的プログラミングを用いて,曲線的に出版者・時代ごとの専用ルビ除去式の生成を行う.In National Diet Library, books which are possessed in library as "the digital library from meiji era" are open to the public on WEB. Since these are shown as image data and cannot search using document contents, an automatic text conversion is needed. However, ruby is a disturbing text conversion. Since existing techniques of linearly removing ruby had developed for books of the current standard, the techniques are inapplicable to early-modern Japanese books, which have a specific characteristic different from characters of current books. In this paper, we propose a method to remove ruby from early-modern Japanese books using Genetic Programming., 2012年11月29日, 2012, 17, 1, 6
日本語, 電子情報通信学会技術研究報告. CPSY, コンピュータシステム, 一般社団法人電子情報通信学会, グラフ解析ワークロードのキャッシュ適合性, 田邊 昇; 冨森 苑子; 高田 雅美; 城 和貴, Graph500ベンチマークはビッグデータ解析のベンチマークとも言われており,近年注目を集めている.本報告ではGraph500ベンチマーク課題の疎行列およびフロリダ大学疎行列コレクションを用い,グラフ解析ワークロードに対するメモリアクセス列のキャッシュメモリへの適合性を,空間的局所性の観点から解析した.その結果,Graph500の疎行列の空間的局所性は極めて低く,疎行列ベクトル積(SpMV)を行う場合にはベクトルのアクセスの際にキャッシュラインには平均して1個程度しか有効なデータは載っていないことが判った.データサイズの大きさとランダム性のため,キャッシュアウェアなソフトウェア最適化が困難である.ゆえに,Graph500リストで優位に立つには,1桁程度の計算時間短縮が期待できるメモリ側のGather機能が有望である., 2012年10月05日, 112, 237, 67, 72
日本語, 年会論文集, 日本教育情報学会, 2PC1 教材開発と知的財産権(教材開発,一般研究,教育情報と人材育成~未来を育む子供たちのために~), 吉井 直子; 篠澤 和恵; 高田 雅美; 城 和貴, 算数・数学で学習する図形(幾何分野)の理解を高めるための教材開発を行い,2013年3月に特許権を取得した.そして,開発した教材を用いて3歳児から5歳児を対象に「算数セミナー」を開催している.本稿では,教育実践と指導用テキスト開発の報告を行うとともに,知的財産権の重要性について述べる., 2015年, 31, 264, 265
日本語, 研究報告数理モデル化と問題解決(MPS), 定在波レーダを用いた距離画像の構築, 張 菁; 梅田 智広; 高田 雅美; 城 和貴, 本研究では,定在波レーダで得られた情報から 3D サーフェイスモデルを構築する.定在波レーダは,非接触で対象物までの距離を測定することができる.定在波レーダを用いて取得した距離情報を処理対象として最適組み合わせアルゴリズムを適用し,その 3 次元距離情報を 2 次元座標に変換し,ドロネー三角分割法を用いて 3 D サーフィスモデルの構築を目指す.In this research, we construct a 3D surface model, which is translated in information of standing wave radar. The standing wave radar can measure distance to an object without contact. In the information of the standing wave radar, an optimal combination algorithm is adopted. Then, once the three-dimensional information is transformed into the two-dimensional coordinates, a 3D surface model is constructed by using the Delaunay triangulation., 2014年02月24日, 2014, 19, 1, 6
日本語, ハイパフォーマンスコンピューティングと計算科学シンポジウム論文集, 次世代HPC・Enterprise統合基盤のためのアプリ上の共通の特徴, 田邊昇; 冨森苑子; 高田雅美; 城和貴, 2013年12月31日, 2014, 43, 43
日本語, 先進的計算基盤システムシンポジウム論文集, Xeon
(R) Phi
TMのGather命令とHMC内のGatherの差異, 冨森苑子; 田邊昇; 高田雅美; 城和貴, 2013年05月15日, 2013, 126, 127
日本語, ハイパフォーマンスコンピューティングと計算科学シンポジウム論文集, Graph500競争を独走するための未開拓領域の予備評価, 田邊昇; 冨森苑子; 高田雅美; 城和貴, 2013年01月08日, 2013, 85, 85
日本語, 研究報告バイオ情報学(BIO), 近代デジタルライブラリーテキスト化支援のためのポータルサイトの設計, 黒田 佳世; 榎本 友里枝; 高田 雅美; 城 和貴, 本稿では,本の見開き画像をテキスト化する際に生じる誤認識を修正する機能をもつポータルサイトを設計する.対象とする画像は,国会図書館が所有する近代書籍のデジタル画像とする.このデジタル画像に対して,近代書籍専用の活字文字認識を適用することによりテキスト化は可能であるが,誤認識が生じる.これを修正するために,デジタル画像とテキスト文書を用いる.提案するポータルサイトでは,テキスト文書と画像データを対応させるメタデータを作成することによって,テキストに対応する画像を表示させることを可能にする.In this paper,we present a design of a portal site which has functions for correcting erratums.These erratums are caused when the facing images,which are owned by Digital Library from Meiji Era in National Diet Library,are transformed into text documents.Although those images can be transformed by using an OCR,which is specialized in early-modern printed books,the OCR infrequently cause erratums.So,to correct them, we take by means of those images and text documents. In this case,we make the metadata by which text can be corresponded to image ., 2010年12月09日, 2010, 35, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 情報処理学会, CUDAに適用したループ再構築手法の評価, 小郷 絢子; 高田 雅美; 城 和貴, 本縞では,既存のプログラムを高速化するために,GPGPU (General Purpose Computing on Graphic Processing Unit) のためのループ再構築手法を適用し,評価する.GPGPU で用いる GPU (Graphic Processing Unit) として,tesla C1060 を採用する.tesla C1060 では GPGPU 用開発環境である CUDA が使用可能である.本縞において,ループ再構築手法例としてストリップマイニングとループ交換を取り上げる.これらのループ再構築手法に対してサンプル・プログラムを適用した際の性能評価を行う.その結果,ストリップマイニングでは分割数の変化によってループアンローリングが適用され,実行時間に影響がでることが分かった.一方ループ交換では,データへのアクセス順序を考慮することによって高速化が可能であることが判明した.これらの結果を踏まえ,実プログラムにおける性能を評価するために,k-means 法にストリップマイニングを,メディアンフィルタにループ交換を適用する.In this paper, we improve and evaluate loop transformation methods in GPGPU (General Purpose Computing on Graphic Processing Unit) for speedup of existing programs. To adopt GPGPU, we use tesla C1060, which is known as a GPU (Graphic Processing Unit). In tesla C1060, CUDA, which is one of development environments for GPGPU, can be used. In this paper, we target strip mining and loop interchange, which are loop transformation methods. To evaluate these loop transformation methods, we validate the performance of these loop transformation methods using simple sample programs. As the result in strip mining, execution time is effected by loop unrolling, which is changed at division number. On the other hand, in loop interchange, through access order to data is attentioned, speed-up can be done. To evaluate the performance in existing programs, we adopt strip mining to k-means and loop interchange to median filter in the light of the results., 2010年09月21日, 2010, 6, 1, 6
日本語, 研究報告ハイパフォーマンスコンピューティング(HPC), 情報処理学会, 機能メモリとGPUのPCI express接続によるヘテロ環境における超大規模疎行列ベクトル積の性能予測, 小川 裕佳; 田邊 昇; 高田 雅美; 城 和貴, 本報告では,疎行列ベクトル積のベクトルがデバイスメモリに入りきらないほど大きな問題向けに,GPU が PCI express 越しに Gather 機能を有する大容量機能メモリをアクセスするシステムを用いた並列処理方式を提案する.フロリダ大の疎行列コレクションを用いて提案方式の性能評価を行った.その結果,間接アクセスの直接アクセス化により,単体性能は既存研究の最大 4.1 倍に向上した.GPU 内キャッシュが溢れる心配も無い.GPU 間通信を完全に排除可能にした構成によりスケーラビリティは保証されており,PCI express のバースト転送バンド幅で制約される単体性能にノード数を乗じたものが並列実効性能となる.In this report, we propose a parallel processing strategy for huge scale sparse matrix-vector product whose vector cannot be held on a device memory. The strategy uses a system connected by PCI express with GPUs and functional memories with gather function. We evaluate the performance of proposed strategy with University of Florida Sparse Matrix Collection. The result shows the 4.1 times acceleration over the existing performance record with a GPU in the maximum case. There is no risk of performance degradation by overflowing cache capacity on GPU. Because of the architecture without inter-GPU communications, scalability is guaranteed. Therefore, parallel effective performance is the product of number of nodes and single GPU performance limited by burst transfer bandwidth of PCI express., 2010年07月27日, 2010, 20, 1, 9
日本語, 研究報告数理モデル化と問題解決(MPS), CUDAを用いたGROMACSの高速化, 畑 典子; 高田 雅美; 城 和貴, 本稿では,GPU のための最適化を施すことによる GROMACS の性能改良について述べる.その最適化には CUDA を用いる.GPU を用いた GROMACS の実行については,GROMACS-OpenMM が既に開発されている.しかしながら,ベンチマークの結果,GROMACS-OpenMM による実行した際,CPU のみで GROMACS を実行した時と比べ,ほとんど実行時間が変わらないシミュレーションや,むしろ遅くなるシミュレーションも存在する.そこで,GROMACS-OpenMM の改良を行う.GROMACS は大規模なプログラムである.そこで,本稿では,関数やメモリアクセスの最適化に焦点を当てる.In this paper, we present some performance improvement of GROMACS for GPUs, which is a well-known application for analysis of molecule, with several optimizations. The optimizations are applied for CUDA. An exsisting GROMACS for GPU, GROMACS-OpenMM has already been developed. However, by our benchmarking results, the execution time of some sample programs of GROMACS-OpenMM are nearly equal or longer than that in GROMACS on a CPU. Therefore, GROMACS with GPU should be improved. Since GROMACS consists of huge number of modules, in this paper, we focus on function calls and memory access., 2010年07月05日, 2010, 10, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 情報処理学会, リアルタイムステレオQueryballの開発, 勅使河原 佑美; 石川 千里; 高田 雅美; 城 和貴, 3 次元空間内のオブジェクトを理解するための機能の 1 つとして Queryball がある.Queryball はオブジェクトに半透明の球体を重ね合わせることにより,オブジェクトと球体が重なる部分に対して問合せることができ,オブジェクトの表示方法を変えたり,様々な方向からオブジェクトを視認したりできる.しかし,現段階の Queryball のボリュームレンダリングでの表示速度はオブジェクトデータのサイズが大きくなるほど遅くなり,実用には向かない.そこで,Queryball をリアルタイムで実行するために,ボリュームレンダリング部分に GPGPU を用いることによって,高速化を図る.また,オブジェクトをよりリアルに理解するために 3 次元立体視を行う.Queryball is a tool to understand objects in three-dimensional space. By overlapping a translucentglobe with 3D object, it can quiry to the part where the object and the globe come in succession. It can change the display method with visually checking according to various directions. However, the display speed the volume rendering of Queryball to the present stage slows the large size of the object data, and is not for practical use. Therefore, performance improvement by applying GPGPU to the volume rendering part in real time is necessary. In addition, the system offers three-dimensional stereoscopic vision for user to understand the object more realistically., 2010年02月25日, 2010, 31, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 情報処理学会, ビデオシーケンスを利用したVR環境マッピング用テクスチャ及び空間の生成, 吉田 幸代; 高田 雅美; 城 和貴, リアルな建物のモデルの作成はサイバーシティアプリケーションの実装において,最も重要なタスクの 1 つである.本稿では見え方を重視し,且つユーザの位置を任意に変更できる臨場感ある屋内空間モデル作りについて考える.具体的には、ユーザの位置によって表示するテクスチャを動的に変更する手法を提案する.既存の見え方を重視した手法では、全方位カメラで撮影した写真をモザイキングし.1 枚の大きなパノラマ画像を用意し,半球にマッピングすることでサイバースペースを作成している.この既存の手法は.大抵街等の建物の外観を対象としており,広い空間作りに適している.廊下といった.屋内の狭い空間作りには不向きである.そこで本稿ではホモグラフィや SIFT を使用し,臨場感を維持しつつ,特に廊下といった狭い空間作りの手法を考えた.ユーザが空間内で移動したとき,ユーザの位置に合致したテクスチャに切り替える.臨場感を維持し,自由に空間内を移動できるようにもする.Appearance of Three-dimensional (3D) building model is one of the most important components in a cyber city implementation and application. In this paper, we propose an approach for generating more freely and higher quality appearance of cyber city by using digital Video sequences than existing ones. This technique is updated depending on positions of the virtual view, and maps texture. In existing techniques, a large panoramic image is given by mosaicing all-directional camera.since the object of the techniques is large space such as city scenes, the techniques are unsuitable for small space such as indoor corridor. Therefore, the proposed a system aims to cyber city, which create automatically, for small space using homography, Scale Invariant Feature Transform (SIFT) and Video sequences. When users change their location in VR space, the system change textures newly. The system keeps realistic sensation and free from location's limits., 2010年02月25日, 2010, 26, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 情報処理学会, 収集方法の異なるウェブアーカイブの相互利用, 梅山 章子; 高田 雅美; 城 和貴, 本稿では,国立国会図書館が運営しているウェブアーカイブの WARP が抱える問題に対処するため,差分収集による新なサーバも同時に使う手法について報告する.差分収集とは,ウェブサイトの更新が行われる毎にリアルタイムで収集,保存を行い,WARP を補うサーバとなる.このサーバは,ウェブアーカイブ用アプリケーションの Wayback を用いて構築する.また,1 つのウェブアーカイブサーバとして両サーバが相対的に機能するかを確認するため,プロトタイプを設計する.この際,簡易版 WARP の構築,簡易版 WARP と差分収集サーバのインタフェースを実装する.WARP is a Web archive system in National Diet Library based on fixed interval collection. Since WARP does not have the incremental collection method, it is desired that a new incremental Web archiving system collaborates with WARP. Such the incremental collection server can be constructed with Wayback applications. Thus,we design a prototype system using WARP and the Wayback based server. Each server needs to work alone while any query must be recognized by the both servers. We use a simple WARP based server and implement the interface between the simple WARP and the Wayback based server., 2010年02月25日, 2010, 32, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 情報処理学会, 反射を考慮した日照量計算手法の評価, 寺前 早苗; 石川 千里; 高田 雅美; 城 和貴, 本稿では,レイトレーシング法の光線追跡を利用した日照量算出の一手法を,実際の都市データに適用した実験結果について述べる.各建造物における総照射量や直接光・反射光別の照射回数を 1 時間ごとに 1 日分算出し,太陽方向や建造物の位置関係などによる違いを確認する.また,毎月 1 日における 1 日の総照射量を 1 年分調べることで,季節による日照量の違いについても確認する.なお,実験に使用したデータは,2 値化された地図画像から日照量算出手法に適したデータ形式に変換したものであり,その変換手順についても述べる.In this paper, we report the experimental result where we apply the routine to calculate the amount of sunlight using ray tracing to real buildings data. We calculate the amount of sunlight and irradiation frequency of direct or reflect light every hour, and confirm the defference by sun direction and buildings position. We also calculate them every month, and contrast the defference by season. In addition, we introduce a method that converts binarized map images into the data format for calculation of insolation., 2010年02月25日, 2010, 36, 1, 6
日本語, 研究報告バイオ情報学(BIO), 情報処理学会, タンパク質立体構造の動きに着目した時系列解析, 鎌田 真由美; 戸田 幹人; 高田 雅美; 城 和貴, 本稿では,ウェーブレット変換と特異値分解により熱揺らぎをするタンパク質の時系列データから動きの特徴を抽出できると考え,得られる動きの特徴を用いて骨格原子の集団運動性に関する解析を行う.まず骨格原子の集団運動性と構造間での相関関係について確認し,機能において重要となる 2 次構造に着目し解析を行う.その結果,2 次構造における相関関係や動き方の違いを確認する.To extract motion feature from the temporal sequence data of proteins with heat fluctuations, we present a feature detection method using a wavelet transform and singular value decomposition in this paper. The extracted features are analysed for collective motion of framework atoms. We investigate collectivity of atom motion and correlation between structures, and analyze secondary structures that play a critical role in the function. As the result, we find correlation and difference motion between secondary structures., 2010年02月25日, 2010, 5, 1, 8
日本語, 研究報告ハイパフォーマンスコンピューティング(HPC), 情報処理学会, 機能メモリとCOTSのPCI express接続によるヘテロ環境向けボリュームレンダリングの設計, 芳野 裕子; 田邊 昇; 小川 裕佳; 高田 雅美; 城 和貴, 本報告では,筆者らが提案している可視化システムにおける 「アクセラレータが PCI express 越しに機能メモリをアクセスする」 という概念を推進するために,ボリュームレンダリングに注目した提案システム向けアルゴリズムを検討し,PCI express の特性を評価する.さらに,その環境に適したボリュームレンダリングを C-ray をベースに設計する. C-ray は Cell/B.E. 用に最適化されたボリュームレンダリングの実装例であり,我々の開発しているシステムは PPE が入っていないので,C-ray をそのまま用いることはできない.そのための移植手法の検討が本稿の目的である.In this report, aiming at our concept of "Accelerators access function memory via PCI express ", we investigate a volume rendering algorithm for our system to evaluate the characteristics of PCI-express. In addition, we design a volume rendering algorithm based on the C-ray for the hetero environment. The C-ray is the optimized implementation of volume rendering for the Cell/B.E. Broadband Engine. Since our system does not have PPEs, we can not apply the C-ray to our system. The purpose of this report is to investigate the re-design of the C-ray for our system., 2009年11月23日, 2009, 1, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 情報処理学会, 対話的実時間遠隔可視化システムの構築, 小堀 由貴; 来見 真理; 石川 千里; 高田 雅美; 城 和貴, 今日,医療現場や気象研究などにおいて三次元可視化手法を利用した解析が活用されている.これら三次元可視化では計算量ならびにデータ量が莫大である一方,その中から解析すべき部分を見つけ出すためには対話的操作が重要となってくる.そこで我々は,遠隔にいる複数の使用者がリアルタイムで解析が可能となる対話的実時間遠隔可視化システムの開発に着手している.本稿では,このシステムで利用されるサーバからの画像配信のための手法について提案する.この手法により,サーバから実時間での画像配信が可能となり,サーバとクライアント間のストレスのない対話的操作の実現が期待できる.Nowadays, 3D visualization based analysis is well-known for various fields such as medical front and weather research. Although they are computation and memory intensive, the users require interactive operations such that they detect a part of space inside of the given 3 dimensional data. To deal with the contradictive processes, we start the development of an interactive real-time remote visualization system, which allows users to perform real-time analysis from remote sites. In this paper, we present a live streaming method for real-time remote visualization. Using the method, we get interactive operations between remote users and real-time visualization servers., 2009年09月03日, 2009, 19, 1, 6
日本語, 研究報告ハイパフォーマンスコンピューティング(HPC), 情報処理学会, ランダムアクセス型応用のためのPCI express越しに機能メモリをアクセスするアーキテクチャ, 田邊 昇; 芳野 裕子; 小川 裕佳; 佐々木 愛美; 高田 雅美; 城 和貴, 本報告では,筆者らが提案した可視化装置における 「アクセラレータが PCI express 越しに機能メモリをアクセスする」 という概念を,ポストベクトル機向けに整理する.ボリュームレンダリングを例として提案装置向けアルゴリズムを検討した.さらに,本用途への GPU 利用の可能性や PCI express の特性を評価した.乱数やボリュームレンダリングの履歴 index を用いた配列間接参照実験では、ミッドレンジ GPU は理論ピークの 1.3% の実効バンド幅でデバイスメモリをアクセスした.このバンド幅は PCI express と機能メモリにより置き換え可能と考えられる.よって,本概念はランダムアクセス型応用を加速するアクセラレータ向けのメモリ容量向上策や、ECC 付加による信頼性強化策として有望である。In this report, we organize the concept on our visualization machine as for the post-vector supercomputer. The proposed concept is the architecture with accelerators accessing functional memory through PCI express. We investigate the volume rendering algorithm for the machine based on the concept. Moreover, we evaluate the feasibility of using GPU and the performance of PCI express for this use. In the experiment of indirect array accesses using random index and index trace on volume rendering, a mid-range GPU accessed device memory with 1.3% bandwidth compared to the theoretical peak. This bandwidth seemed to be able to be replaced with PCI express and functional memory. Therefore, this concept is promising for accelerator running applications with random accesses both to overcome limitation of device memory capacity, and to improve robustness of memory using ECC., 2009年07月28日, 2009, 25, 1, 8
日本語, 情報処理学会研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, 芝生用塗料の劣化モデルの構築, 榎本 友理枝; 石川 千里; 高田 雅美; 城和貴, 本稿では,芝生の色の保持と緑色塗料の関係を予測するためのモデルを構築する.モデル構築にあたって,実際に試験塗料を散布させた芝生の劣化変化の解析結果を利用する.このモデルを用いてシミュレーションを行うことにより,芝生の色の持続性と塗料の濃度の関係が分かり,芝の品質に影響を与えない緑色塗料を適切な時期に散布し,常に葉色の良い状態で芝生を維持することが可能になると期待される.By this study, we build a model to predict color maintenance for lawns and relations of the green paint. On the model construction, We use the analysis results of the deterioration change of lawns which get really scatter examination paint. We understand the durability of the color of lawns and relations of the density of paint by simulating with this model. We can scatter the green paint which does not affect the quality of the turf in an appropriate time. And it is always expected with good leaf color and can maintain lawns., 2008年12月10日, 2008, 126, 17, 20
日本語, 情報処理学会研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, 反射を考慮した日照量計算の一手法, 樽井 沙知; 芳野 裕子; 寺前 早苗; 石川 千里; 高田 雅美; 城和貴, 本稿では,レイトレーシング手法の光線追跡を利用して,日照量算出のための計算手法を提案する.この手法では,周囲の建物による太陽光線の遮蔽,反射の影響も考慮に入れたうえで日照量の計算を行う.太陽光発電は,発電設備の設置場所によって得られる発電量が大きく異なるという問題点を持つ.そこで,この手法を適用することにより,発電量をより多く得られる場所を特定することができると期待される.また,大規模データに対する計算になるため,データ量及びメモリ削減を行うためのデータ構造の定義と計算手法についても提案を行う.In this paper, we introduce the calculation routine to calculate amount of sunlight using ray tracing. In this approach, we calculate amount of sunlight to considered influence of cover and reflection. Photovoltaic generation has the problem that the electrical energy obtained by the setting position of photovoltaic facility differ greatly. Then, we expect to pinpoint a place which can obtain more electrical energy by applying this approach. And, because it is the calculation to large-scale data, we introduce the definition of the data structure and calculation approach to reduce data and memories., 2008年12月10日, 2008, 126, 45, 48
日本語, 情報処理学会研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, 多目的に衛星データを利用するためのコンテンツ開発環境, 田中 里奈; 石川 千里; 高田 雅美; 城和貴, 本稿では,衛星データをデータベース化してメタデータを付加することにより様々なクエリに対しての処理を行うことが可能となるようなコンテンツの一部である,リアルタイムに立体視画像を生成し,表示する Web コンテンツの機能を提案する.Web での動作をインタラクティブなものにするために Ajax を用いる.また,立体視はアナグリフ画像を生成して行う.これにより,商用や研究用などの限られた専門的な分野でしか利用されていない衛星データを一般のユーザも利用可能になる.In this report, we generate a stereoscopic vision image in real time that we can handle for various queries by compiling satellite data into a database, and adding meta data to suggest a function of the web content to display. We use Ajax to make interactive movement on the Web . In addition, the stereoscopic vision generatesd by a anagliph image. The satellite data were used so far only in limited technical fields such as the business and study use, but general users can use these contents easily., 2008年12月10日, 2008, 126, 5, 8
日本語, 情報処理学会研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, Flashを用いたWebアルバムの提案と設計, 重見 早南; 石川 千里; 高田 雅美; 城和貴, 本稿では,Flash を用いた Web アルバムの提案と設計について述べる.アルバムの自動生成の作成を行うシステムで使用する,関連画像と関連文書の提示を行うツールの開発を行う.関連画像は,写真の代わりとなる画像を取得するために用いる.しかし,ただ画像を提示するだけでは,ユーザのオリジナルアルバムとは言えない.そこで画像に付随して生年月日などからその当時の流行や音楽等の関連文書を提示することで,オリジナルアルバムを作成することができる.In this paper, we report the design of a Web album using Flash. For the automatic generation of the Web album, tools are developed showing related images and documents. The related images are used instead of photograph. A web album for a person should not be generated just from his/her related images. Therefore, by having related documents such as the fashion or the music of the days of birth, we can make the web album., 2008年12月10日, 2008, 126, 1, 4
日本語, 情報処理学会研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, Webアーカイブにおける差分収集に用いる更新間隔推定手法の開発とその評価, 廣道尚弓; 石川 千里; 高田 雅美; 城和貴, 本稿では,Web ページの更新間隔を推定する手法を開発する.この手法を用いて Web ページを差分収集するデータベースを構築する.これは,Web アーカイブにおいて,Web ページの更新,変更,削除に対応するために用いられる.開発にあたって,Web ページが更新される間隔の分布型を予め仮定する.そのため,ブートストラップ法が適用可能である.In this report, We develop a technique to estimate the update period of a Web page. We build a database collecting a Web page to validate this technique. This is used corresponding to the update, the change and the deletion of the Web page in a Web archive. On development, We assume the distribution type of the interval when a Web page is updated beforehand. Therefore We can apply a boot strap method., 2008年12月10日, 2008, 126, 41, 44
日本語, 情報処理学会論文誌数理モデル化と応用(TOM), 情報処理学会, 「情報処理学会論文誌:数理モデル化と応用」の編集にあたって, 城和貴; 北栄輔(ゲスト・エディタ); 北 栄輔; KITA Eisuke, 2007年12月15日, 48, 19, i, i
日本語, 年会論文集, 日本教育情報学会, エッジ画像における奥行き知覚を利用した教育への一提案 (教育資料・実践(2)), 吉井 直子; 高田 雅美; 城 和貴, 初等中等教育現場では,黒板を使用する機会が多い.黒板にかいたは3次元物体は,エッジ画像に近くなる.そこで,本稿ではエッジ画像を用いて奥行き知覚の能力を調べる.実験結果から回帰分析によって発達モデル式を構築する.さらに,モデル式の精度評価を行い,エッジ画像の奥行き知覚について,男子と女子の間には発達の違いがあることを理論値で示す.最後に,それを補完するために教育場面で活用できる教材を提案する., 2010年08月, 26, 170, 173
日本語, 年会論文集, 日本教育情報学会, 視覚特性(オクリュージョン効果)を利用した教育への試み, 吉井 直子; 石川 千里; 高田 雅美; 城 和貴, 初等中等教育現場では印刷物をよく使用する.ところが,子どもの2次元画像の認識に関する既存研究は少なく,男女別画像別に分析したものは,我々が知る限り存在しない.そこで,奥行き知覚の発達過程について実験を行い考察した.実験結果から,2次元画像の奥行き知覚について男子と女子の間には発達の違いがあること,黒板にかかれた空間図形などのエッジ画像では正解率が低くなることが検証できた.そこで,この問題を補完する一例としてオクリュージョン(重なり)効果を採り上げる., 2009年08月, 25, 156, 159
日本語, 研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, 近代書籍に特化した多フォント活字認識法, 芦田 尚美; 高田 雅美; 木目沢 司; 城 和貴, 国立国会図書館は,所蔵する明治から大正期にかけての近代書籍を近代デジタルライブラリーとしてWeb上で一般に公開している.公開されている書籍は全て画像としてデータ化されており,全文検索を用いて書籍内容の検索を行うことができないため,早急なテキストデータ化が求められている.しかし,旧字体を多く含み活字の種類が特定できない近代書籍はOCRソフトウェアによるテキスト化を適用できない.そこで,本研究では近代書籍に特化した活字認識の手法を提案する.National Diet Library has a public web site as Early-Modern Digital Library for books in Meiji and Taisho periods. Since the archive is digitalized as images, text search for the books is not applicable and, therefore, required. Conventional OCR systems are not a good tool for such modern books because they have various font sets and most of them are very noizy. In this paper, we propose a character recognition method for early-modern printed books., 2009年02月26日, 2009, 19, 205, 208
日本語, 研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, 地震短期予測研究のための地電流解析ポータルの開発, 豊島 良美; 石川 千里; 高田 雅美; 長尾 年恭; 城 和貴, 本稿では,地電流を用いた解析システムの開発について述べる.このシステムは,解析結果の収集と研究者の支援を目的としている.これらの目的を達成するために必要な機能について検討を行い,その検討を基にシステムを構築する.そして,構築したシステムが機能要求を満たすかの確認を行う.In this paper, we present the development of an analysis system for telluric current data. By using the proposed system, analytical results are collected to support researchers. Necessary functions are examined to construct the proposed system. We also verify that the system satisfies these necessary functions to meet the demands., 2009年02月26日, 2009, 19, 209, 212
日本語, 研究報告数理モデル化と問題解決(MPS), ARを用いた空間認識能力向上のための学習方法, 秦野 真衣; 米澤 朋子; 吉井 直子; 高田 雅美; 城 和貴, 本稿では,子どもの空間認識能力を高めるための AR を用いた新たな学習方法を提案する.AR を用いる利点として,物体を実際に動かすことで子どもでも直感的に,動作と視覚の両方の感覚を使って扱うことができる点がある.また,空中に浮いている立体などの仮想立体でも提示することが可能である.本学習方法は,イメージの中で物体を回転させやすくなるような支援を行っている.学習者は AR 空間の中に表示された物体を別の角度から見る場合を想像し,想像された物体を新たに作成するものである.この学習方法による空間認識能力の学習効果を調べるために,小学生を対象とした実験と実験結果による分析を行っている.In this research, we propose a method to progress spatial reasoning capacity of children with Augmented Reality(AR). AR has two advantages. First, children can use AR tools viscerally with sence of action and vision, since this tool can be moved actually. Second, AR tools with AR can display virtual solids like floating solid. An AR tool, which is based on the proposed method, supports to make image which rotates objects more easily. Learner images rotated objects which was displayed in AR. Then, they make this imaged objects. To evaluate spatial reasoning capacity in the AR tool, we experiment and analyze in regard to elementary school children., 2012年02月23日, 2012, 33, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 近代デジタルライブラリーの文字切り出しにおける実際的手法, 福尾 真実; 高田 雅美; 城 和貴, 本研究では,近代デジタルライブラリーが所蔵する画像データから上手く文字を切り出す実際的手法の開発を行う.国立国会図書館では近代デジタルライブラリーとして,所蔵する書籍を Web 上で一般公開している.これらは,画像データとして公開されており,文書内容を用いた検索が行えないため,早急なテキスト化が求められている.そのため,近代書籍に特化した多フォント漢字認識手法が提案されている.しかし,ルビが振られた書籍からは上手く文字が切り出せず,認識ができない.そこで本稿では書籍の本文からルビを取り除く手法を開発する.In this research, we develop a practical technique to clip the kanji character well from the image data that the digital library from meiji era houses. The national diet library in Japan is opened to the public as the digital library from meiji era on the Web. There are shown as image data. Since it is impossible to perform full text search, it should be converted to text data. Therefore, it has been proposed the multi-fonts kanji character recognition method for early-modern Japanese printed books. Kanji characters with rubi occur that the kanji character clipping and recognition are badly constructed. In this paper, we propose a technique to remove the rubi from body of the book., 2012年02月23日, 2012, 34, 1, 6
日本語, 研究報告バイオ情報学(BIO), 未病の認識・改善に有効なストレスオントロジーの設計, 洞渕 彩未; 高田 雅美; 森本 恵子; 城 和貴, 本研究では,オントロジー技術を用いて,ストレスに関する情報の関係性を示したストレスオントロジーを設計する.開発システムでは,センサデバイスで取得した生体情報とオントロジーを利用することで,各ユーザ専用の未病の認識と改善に効果的な情報を提供する.開発するオントロジーでは,Web 上に溢れている情報を収集し,機械が処理できる形式で管理する.本論文では,設計したオントロジーと生体データを用いた処理の評価と,矛盾を検出し最適化する手法についても考察する.In this paper, we show the design of the stress ontology which describes the information of stress affection. Using the developed system, we provide the information which is effective for the presymptomatic various diseases. Using the ontology, we collect the information from the web to re-construct the ontology. In this paper, we explain the design of our health care ontology and its practical use for the real. The structure optimization and the removal of contradiction of the proposed ontology is an important research topics., 2011年11月24日, 2011, 26, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 異なるプラットフォームで効果的な3Dモデルを表示するためのポリゴンリダクションの方法, 水野 志織; 奥本 萌; 高田 雅美; 城 和貴, 本研究では,効果的に表示できる最小のポリゴン数の定義を行う.近年,文化財のデジタルデータによる保存が行われている.しかし,ポリゴンのデータ量は莫大で,高い処理能力を持った計算機が必要とされる.一般公開を行う場合,処理能力の低い計算機への適用がなされるべきである.そのため,ポリゴンリダクションを用いて莫大なデータ量を少なくする必要がある.この際,適切なポリゴン数に削減すべきである.この数を定義するために,評価実験を行う.この実験結果として,3次元モデルを表示するピクセル数の50倍が最適であることが分かる.In our research, we define the minimum number of polygons for effective rendering. Recently cultural property is preserved in digital data. Since polyangular data is large, a high performance computer is needed. If it is shown to the public, a low performance should be targeted. Therefore, polygons are reduced to a valid number. We try to define the minimum number. As a result, we find that the number of polygon should be 50 times as many as the number of pixels for rendering 3D models., 2011年11月24日, 2011, 16, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 未病の認識・改善に有効なストレスオントロジーの設計, 洞渕 彩未; 高田 雅美; 森本 恵子; 城 和貴, 本研究では,オントロジー技術を用いて,ストレスに関する情報の関係性を示したストレスオントロジーを設計する.開発システムでは,センサデバイスで取得した生体情報とオントロジーを利用することで,各ユーザ専用の未病の認識と改善に効果的な情報を提供する.開発するオントロジーでは,Web 上に溢れている情報を収集し,機械が処理できる形式で管理する.本論文では,設計したオントロジーと生体データを用いた処理の評価と,矛盾を検出し最適化する手法についても考察する.In this paper, we show the design of the stress ontology which describes the information of stress affection. Using the developed system, we provide the information which is effective for the presymptomatic various diseases. Using the ontology, we collect the information from the web to re-construct the ontology. In this paper, we explain the design of our health care ontology and its practical use for the real. The structure optimization and the removal of contradiction of the proposed ontology is an important research topics., 2011年11月24日, 2011, 26, 1, 6
日本語, 研究報告バイオ情報学(BIO), MKLによる食事画像認識の追試, 内村 麻里奈; 高田 雅美; 城 和貴, 本研究では食事画像認識をする.食事のカロリーなどの情報を必要としているユーザがいる.しかしながら,食事の種類はとても多いため,一般ユーザにとって,正しい食事の情報をえることは困難である.この問題を解決するために,食事画像を使って専門家がアドバイスするシステムがある.ただし,このシステムでは,処理能力に限界がある.そこで,食事画像に対して,画像認識技術を適用することによって,認識の自動化を行う.分類機の1つとして,SVM(Support Vector Machine) がある.この SVM は,入力された画像の特徴を用いて,正しく分類することができる.食事の種類は多いため,複数の特徴で表す必要がある.そのため,これらの特徴を SVM で利用するために,統合しなければならない.このための手法として,MKL(Multiple Kernel Learning) を用いる.この MKL を用いた SVM の性能を調べるために,85 種類の食事画像を用いて,実験を行う.In this paper, we recognize food image. Users need information about food calorie so on. However, it is too difficult to have correct information about food because of variety foods. To solve this problem, the system which specialists advise with food image has been developed. However, this system has a limitation in capacity. Therefore, food images should be recognized automatically by using image-recognition technique. SVM (Support Vector Machine) is one of supervised learning. The SVM can classify correctly with given image features. A number of features are needed to treat so many food. These features have to be combined by using SVM. Consequently, MKL (Multiple Kernel Learning) should be adopted. We experiment with 85 categories of food images to find out quality of SVM using MKL., 2011年11月24日, 2011, 17, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), MKLによる食事画像認識の追試, 内村 麻里奈; 高田 雅美; 城 和貴, 本研究では食事画像認識をする.食事のカロリーなどの情報を必要としているユーザがいる.しかしながら,食事の種類はとても多いため,一般ユーザにとって,正しい食事の情報をえることは困難である.この問題を解決するために,食事画像を使って専門家がアドバイスするシステムがある.ただし,このシステムでは,処理能力に限界がある.そこで,食事画像に対して,画像認識技術を適用することによって,認識の自動化を行う.分類機の1つとして,SVM(Support Vector Machine) がある.この SVM は,入力された画像の特徴を用いて,正しく分類することができる.食事の種類は多いため,複数の特徴で表す必要がある.そのため,これらの特徴を SVM で利用するために,統合しなければならない.このための手法として,MKL(Multiple Kernel Learning) を用いる.この MKL を用いた SVM の性能を調べるために,85 種類の食事画像を用いて,実験を行う.In this paper, we recognize food image. Users need information about food calorie so on. However, it is too difficult to have correct information about food because of variety foods. To solve this problem, the system which specialists advise with food image has been developed. However, this system has a limitation in capacity. Therefore, food images should be recognized automatically by using image-recognition technique. SVM (Support Vector Machine) is one of supervised learning. The SVM can classify correctly with given image features. A number of features are needed to treat so many food. These features have to be combined by using SVM. Consequently, MKL (Multiple Kernel Learning) should be adopted. We experiment with 85 categories of food images to find out quality of SVM using MKL., 2011年11月24日, 2011, 17, 1, 6
日本語, 研究報告バイオ情報学(BIO), 異なるプラットフォームで効果的な3Dモデルを表示するためのポリゴンリダクションの方法, 水野 志織; 奥本 萌; 高田 雅美; 城 和貴, 本研究では,効果的に表示できる最小のポリゴン数の定義を行う.近年,文化財のデジタルデータによる保存が行われている.しかし,ポリゴンのデータ量は莫大で,高い処理能力を持った計算機が必要とされる.一般公開を行う場合,処理能力の低い計算機への適用がなされるべきである.そのため,ポリゴンリダクションを用いて莫大なデータ量を少なくする必要がある.この際,適切なポリゴン数に削減すべきである.この数を定義するために,評価実験を行う.この実験結果として,3次元モデルを表示するピクセル数の50倍が最適であることが分かる.In our research, we define the minimum number of polygons for effective rendering. Recently cultural property is preserved in digital data. Since polyangular data is large, a high performance computer is needed. If it is shown to the public, a low performance should be targeted. Therefore, polygons are reduced to a valid number. We try to define the minimum number. As a result, we find that the number of polygon should be 50 times as many as the number of pixels for rendering 3D models., 2011年11月24日, 2011, 16, 1, 6
日本語, 研究報告計算機アーキテクチャ(ARC), 不規則型応用を加速するメモリアクセラレータ - Exa FLOPSマシンの文脈から, 田邊 昇; Boonyasitpichai Nuttapon; 中條 拓伯; 小郷 絢子; 高田 雅美; 城 和貴, Byte/FLOP を現在より下げざるを得ない Exa FLOPS 級マシンにおいて,国内では階層キャッシュが重要視されている.一方,不規則型応用は必要な Byte/FLOP が大きく,かつ,文科省の重点指定アプリの中でも大きな比率をしめており,日本の Exa FLOPS 級マシンの文脈において対応が重要である.本報告では Gather 機能を有するメモリシステムについて上記の文脈上で考察する。その判断材料として現状の GPU のキャッシュヒット率と,キャッシュを前段に併用した Gather 機能を有するメモリシステムについて疎行列ベクトル積の性能評価を行なった.その結果,キャッシュによる性能向上は限定的で,その容量の 10 倍程度のベクトルサイズまでで頭打ちとなった.不規則型応用に対して再利用性に高速化原理をおくキャッシュに頼りきることは危険であり,Gather 機構が重要であることが確認された.In Japan, hierarchical cache is currently given high priority for memory system of Exa FLOPS machine whose Byte/FLOP ratio must be smaller than current supercomputers. On the other hand, irregular applications require higher Byte/FLOP ratio. Since these occupy a large part of MEXT-selection of focused applications, supporting irregular applications is important in the context of Japanese Exa FLOPS machine. In this report, a memory system with gather functions is reconsidered in the context of Exa FLOPS machine. We explored hit ratio of a current GPU's cache and performance of memory system with gather function and its preliminary stage cache for sparse matrix vector multiplication. As a result, performance gain by the additional cache is limited and disappeared for vector with about 10times higher size than the cache capacity. It is confirmed that the importance of gather function for irregular applications is higher than that of cache whose acceleration principle is based on reusability., 2011年11月21日, 2011, 14, 1, 8
日本語, 研究報告数理モデル化と問題解決(MPS), e-口コミを活用したe-learning教材の開発とその評価, 松本 多恵; 高田 雅美; 城 和貴, 本稿では,e-口コミを用いた新たな学習支援手法を提案する.この手法の目的は,1)高いドロップアウト率,2)モチベーション維持の困難さ,3)少ないインタラクティブ性という 3 つの e-learning が抱える問題を解決することである.提案する学習支援手法は,従来の学習支援と大きく異なり,メンターなどの学習支援者を必要とせず,学習者自らが意欲的・主体的に学習をする仕掛けを与えるものである.我々は,この e-口コミを発生させるために,消費者購買行動モデルの 1 つである AISAS に着目する.このモデルを応用し,マンガ教材を利用した e-learning 学習教材の開発を行う.開発された教材が e-learning が抱える問題を解決する方法として有効であることをアンケートから示す.In this paper, we propose an e-learning support method using e-WOM (e-Word of Mouth) for three problems in e-learning: 1) learners drop out at high rate 2) it is difficult to maintain motivation 3) it has few interactive. The proposed method is a lot different from existing methods. The proposed method does not need support persons such like a mentor, and set up a case that learners work aspiringly and actively. To go on e-WOM, we adopt AISAS, which is a model for buying behavior. We put AISAS into the development of the e-learning support method using comic learning tools. To validate the method, we have questionnaires., 2011年09月08日, 2011, 27, 1, 6
日本語, 研究報告ハイパフォーマンスコンピューティング(HPC), メモリアクセラレータで強化したGPUのCG法による評価, 小郷 絢子; 田邊 昇; 高田 雅美; 城 和貴, 本報告では,係数行列や解ベクトルが GPU 上に載りきらないほど大きな連立一次方程式を共役勾配法 (CG法) で解く際に,メモリアクセラレータの利用を提案する.提案アクセラレータは GDDR5 ポートなどに接続され,デバイスメモリの厳しい容量制約を緩和するとともに,Gather 機能によりキャッシュや GDDR 系メモリが苦手とする間接アクセスを連続アクセス化する.本報告では,フロリダ大学の疎行列コレクションを用いて提案方式の性能評価を行った.その結果,テクスチャキャッシュが効くような小さな行列でも,単体性能はテクスチャキャッシュを用いる既存手法の 1.05~2.01 倍に向上した.従来手法は行列サイズを大きくした時,GPU 内キャッシュのヒット率が低下し,性能低下する.解ベクトルがデバイスメモリ容量を超えると PCI express を通過する通信により,さらなる性能低下が予想される.それに対し,本手法はそれらの心配が無い.In this paper, we propose the use of memory accelerator to solve systems of liner equation, which coefficient matrices and solution vector unable to be had on board by CG method. In the case of existing method, when the size of matrix is increase, performance decrease because of hit probability on GPU cache decrease. We predict that performance decrease because of access of exceed access, the case solution vector is over device memory capacity. In contrast, there is no risk, thanks to this method. Proposed accelerator is connected to such as GDDR5 port, it ease severe capacity limit, and make indirect access which is unsuitable for cache and GDDR memory into direct access using gather function. In this paper, we evaluate the performance of proposed strategy with University of Florida Sparse Matrix Collection. The result showed from the 1.05 to the 2.01 times acceleration over the case of existing performance record with the texture cache, even if small matrix which has effect of texture cache., 2011年07月20日, 2011, 15, 1, 10
日本語, 研究報告ハイパフォーマンスコンピューティング(HPC), 情報処理学会, GPUと拡張メモリによる疎行列ベクトル積性能の行列形状依存性軽減, 田邊 昇; Boonyasitpichai Nuttapon; 中條 拓伯; 小川 裕佳; 高田 雅美; 城 和貴, GPUのメモリシステムにおける不連続アクセスにおける問題を解決するために筆者らは先行研究でScatter/Gather機能を有する拡張メモリシステムを提案した。疎行列ベクトル積に対する評価も行ない,提案拡張メモリシステム側の疎行列ベクトル積に対するスループットをPCI expressのバンド幅以上にできることを示した。しかし,疎行列ベクトル積に関しては評価行列が1つのみであった。本報告ではその種類を増やした実験によって,提案方式の性能の行列形状依存性について評価した.その結果,提案方式は行列の形状に性能が鈍感で,高い実行性能を安定して提供できることがわかった.In order to overcome the problems of discontinuous accessing in memory systems of various processors, we have proposed an extended memory system which has an additional function of scattering and gathering. In this paper, we have confirmed the assumption on feasibility of throughput of the memory system is correct or not with a simulator based on DRAMsim2 with various matrices. As the result, we have confirmed proposed system has good stability for matrix figure., 2011年03月08日, 2011, 11, 1, 8
日本語, 研究報告数理モデル化と問題解決(MPS), 情報処理学会, GPUを用いた複数顔検出の高速化に関する一手法, 久保 友理恵; 高田 雅美; 城 和貴, 本稿では,複数の人物の顔を高速に検出するプログラム開発のために,GPU (Graphic Processing Unit) を用いた高速化の一手法を検討する.GPU として,NVIDIA 社の tesla C1060 を採用する.tesla C1060 では,GPGPU (General Purpose Computing on Graphic Processing Unit) 用開発環境である CUDA が使用可能である.顔検出には高速に検出可能な Haar-Like 特徴を用いた AdaBoost ベースの識別器を利用する.検討手法では,入力画像上をランダムに選んだ複数の位置から走査を開始し,Haar-Like 特徴演算を GPU で行うことによって,CPU での実装より高速な処理を可能にする.In this paper, we present some optimization techniques for multiple faces detection using GPU. As the target platform, we use tesla C1060 which is widely know as a high-end GPU. We use CUDA as the development environment for GPGPU. To detect face, we use a classifier based on AdaBoost with using Haar-Like features which can detect the face positions very quickly. In this case, we perform the Haar-Like feature calculation on tesla C1060 to demonstrate a faster implementation rather than CPUs., 2011年02月28日, 2011, 21, 1, 6
日本語, 年会論文集, 日本教育情報学会, 幼児の大画面3DCG体験学習会の報告と今後の課題 (日本教育情報学会 第27回年会 教育情報のイノベーション--デジタル世代をどう導くか) -- (初中等教育での学習教育支援・実践と情報教育), 吉井 直子; 篠沢 和恵; 高田 雅美; 城 和貴, 幼稚園教育は,新幼稚園教育要領にも記されているように環境を通して行う活動を基本としている.本稿では,我々の実施している,幼児を対象とした「大画面3DCG体験学習会」について報告する.これは,新たな環境となりつつある3DCGの最新技術に幼児が触れ,サイエンスやテクノロジーに関心を持つきっかけを与えることを目的とするものである.アンケート結果から,2D作品鑑賞と比較すると,3DCG体験の方がより幼児たちは興味や関心を持つという傾向が示唆される.また3年間実施した経験から,今後幼稚園教育で利用するための課題について述べる., 2011年08月, 27, 150, 153
日本語, 年会論文集, 日本教育情報学会, 3B1 エッジ画像における幼児期の奥行き知覚の特徴(一般研究 初中等教育の情報教育1,教育情報のイノベーション~デジタル世代をどう導くか~), 吉井 直子; 高田 雅美; 城 和貴, 幼児が効果的に形の学習を行うためには,適切な教材が必要である.そこで,本稿では,物体の輪郭を取り出すエッジ抽出が施されたエッジ画像を用いて実験を行い,幼児の奥行き知覚に関する特徴を考察する.実験結果から,5歳児男子は5歳児女子より奥行きをよく知覚していること,5歳児女子は個人差が大きく,平行に並ぶものより重なりがあるものをよく知覚できるということが検証できた., 2011年, 27, 258, 259
日本語, 研究報告数理モデル化と問題解決(MPS), 視覚障害者のための墨字学習支援システムの開発と評価, 松本 多恵; 伊奈 諭; 高田 雅美; 城 和貴, 情報通信機器を活用したコミュニケーションは視覚障害者と晴眼者との有効な情報伝達手段であるが,視覚障害者には大きな壁が存在する.それは 『漢字仮名交じり文』 を作成することである.視覚障害者が使用する点字には,ひらがな,カタカナ,漢字の区別がなく,表音文字だけをどんどん読み進んで文字の意味を理解するため, 『墨字』 という大きな壁にぶち当ってしまう.これまでも視覚障害者向けに点字ワープロ等が開発されてきたが,これらの機器を有効に活用するためには,視覚障害者の墨字リテラシー強化が必要不可欠である.そこで我々は,筆順に着目し,触覚と音声ガイドによる自学自習および対面指導を支援するシステムを開発し,視覚障害者を被験者とした評価実験を行った.We develop a Sumiji learning support system for visually impaired people. The system is based on hand writing using tactile stroke order to be displayed. In order to validate the system, we perform experimental evaluation of the system with visually impaired people. As a result, it is clarified that our system is very effective for the blind to learn Sumiji. Its hand writing is accompanied by shortening of acquisition time and the improvement in the rate of learning is compared with the conventional face to face instruction., 2012年09月12日, 2012, 1, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 古典油彩画グレーズ技法シミュレーションの高速化, 篠澤和恵; 楜沢順; 高田雅美; 城和貴, コンピュータ上での絵具の発色シミュレーションは,グラフィックソフトウェアへの応用が期待できる.古典油彩画における 「グレーズ」 は,薄く溶いた透明な絵具の層を重ねる技法であり,混色で表現できない色調を,重色によって表現するものである.その発色は,放射伝達方程式を解くことによって求められる.しかし従来手法は,異なる絵具でグレーズを行う方法が示されておらず,高精度な解を得るには計算量が大きくなるため,グラフィックソフトウェアへの応用には不向きである.本研究では,まず従来手法を異なる絵具の層に対応できるように拡張する.さらに,係数行列を分解し,刻み幅を固定することにより,高精度な解を小さい計算量で得ることが可能な数値解法を提案する.また,実際の油絵具のデータを用いて数値実験を行い,高精度な解が,従来手法の約 4 % ~ 7 %の計算時間で求められることを示す., 2012年09月12日, 2012, 7, 1, 6
日本語, 研究報告数理モデル化と問題解決(MPS), 同一出版者の近代書籍に対する漢字認識評価, 福尾 真実; 高田 雅美; 城 和貴, 本研究では,同一出版者の近代書籍に対する漢字認識実験を行う.国立国会図書館では,近代デジタルライブラリーとして,所蔵している書籍を Web 上で公開している.これらは画像データとして公開されており,文書内容を用いた検索は行えない.そのため早急なテキスト化が求められている.書籍のスキャンは,見開きの状態で行われている.ゆえに,テキスト化を行う際,文書の傾き補正を行う必要があるが,本稿では新たな角度補正の手法を提案する.また,出版者で使用しているフォントが異なっており文字認識が困難となっている.そこで,本稿では同一出版者の近代書籍に対し漢字の認識実験を行う.In this research, we perform the kanji recognition experiment for the modern book of the same publisher. In National Diet Library, a book possessing as the digital library from meiji era is shown on Web. These are shown as image data and cannot perform the search using document contents. Therefore it is demanded an immediate text becoming it. The scan of the book is carried out in the state of the spread. In addition, fonts using in publishers are different, and character recognition becomes difficult. When we become the text of the book, it is necessary to make the degree of leaning revision of the document, but suggests technique of new angle revision in this report. Therefore we perform the recognition experiment of the kanji for the modern book of the same publisher in this paper., 2012年09月12日, 2012, 26, 1, 6
日本語, 研究報告ハイパフォーマンスコンピューティング(HPC), 疎行列のキャッシュへの適合性分類に関する予備評価, 冨森 苑子; 田邊 昇; 高田 雅美; 城 和貴, エクサスケールマシンは複雑なメモリシステムとなることが予想されている.同マシンへの適用を視野に入れた疎行列ライブラリの実現に向け,本報告では疎行列のキャッシュへの適合性分類に資する疎行列の特性に関する新しい指標として 「列インデックス列の空間的局所性」 を提案する.さらに,入力疎行列および Fold 法前処理後の提案指標の値をフロリダ大学の疎行列コレクションを用いて評価した.その結果,疎行列ベクトル積処理性能と L1 キャッシュヒット率と新指標の間には有意な相関関係があることが確認できた.よって,従来から指摘していた行列サイズと併せ,本指標をアプリ固有の最適化を避けたメモリアクセス機構や前処理アルゴリズム自動選択の指標の一つとする.In Japan, memory system of ExaFLOPS machines is expected more complex. In this paper, we propose a new characteristic of sparse matrices about spatial locality of row-index sequences in order to classify suitability for cache memory systems. Moreover, we evaluate proposal characteristic of input matrices and pre-processes (folding). Test matrices are chosen from University of Florida Sparse Matrix Collection. As a result, it is confirmed that there are significant correlations between performance of Sparse Matrix-Vector Product(SpMV) and the general purpose cache (L1) hit rate. Therefore, our characteristic is suitable for auto-tuning pre-processes and memory access mechanisms to avoid application specific optimization in conjunction with matrix size (the number of rows)., 2012年07月25日, 2012, 17, 1, 7
日本語, 先進的計算基盤システムシンポジウム論文集, 大規模疎行列やメモリシステムの特性を考慮した高速な汎用疎行列ライブラリ実現に向けて, 冨森苑子; 田邊昇; 小郷絢子; 高田雅美; 城和貴, 2012年05月09日, 2012, 65, 66
日本語, ハイパフォーマンスコンピューティングと計算科学シンポジウム論文集, Gather機能を有するメモリアクセラレータの疎行列計算への応用, 田邊 昇; 小郷 絢子; 小川 裕佳; 高田 雅美; 城 和貴, 2012年01月17日, 2012, 32, 41
英語, 情報処理学会論文誌数理モデル化と応用(TOM), 一般社団法人情報処理学会, A Combined Data and Program Partitioning Algorithm for Distributed Memory Multiprocessors, TSUNEO NAKANISHI; KAZUKI JOE; CONSTANTINED.POLYCHRONOPOULOS; AKIRA FUKUDA, In this paper we propose an algorithm to perform data partitioning and program partitioning simultaneously on an intermediate representation for parallelizing compilers which we have proposed the Data Partitioning Graph. Conventional and therefore conservative parallelizing compilers usually activate program partitioning prior to data partitioning. However on distributed memory multiprocessors it is quite difficult to partition a program effectively with consideration of data partitioning since communication costs change depending on a data partitioning and distribution decision. The proposed algorithm resolves this conflict by handling these inseparable partitioning problems simultaneously with and A* algorithm.In this paper we propose an algorithm to perform data partitioning and program partitioning simultaneously on an intermediate representation for parallelizing compilers which we have proposed, the Data Partitioning Graph. Conventional and, therefore, conservative parallelizing compilers usually activate program partitioning prior to data partitioning. However, on distributed memory multiprocessors it is quite difficult to partition a program effectively with consideration of data partitioning since communication costs change depending on a data partitioning and distribution decision. The proposed algorithm resolves this conflict by handling these inseparable partitioning problems simultaneously with and A* algorithm., 1999年12月15日, 40, 9, 30, 41
日本語, 情報処理学会研究報告数理モデル化と問題解決(MPS), 一般社団法人情報処理学会, リリース・コンシステンシ・モデルとその実現の形式的仕様記述について, 高田司郎; 田口研治; 城和貴; 福田晃, 我々は,分散共有メモリシステムの振る舞いを定義したメモリ・コンシステンシ・モデルとその実現の形式的な仕様記述とその検証の研究を行っている.我々は,仕様記述言語Zとvalue-passing CCSを統合した形式的技法をコーザル・メモリ・コンシステンシ・モデルに適用した結果,そのモデルと実現の形式的仕様記述,および,その実現の検証がこの技法で一貫して行えることを確認した.そこで,本稿では,リリース・コンシステンシ・モデルに同様の形式的技法を適用して,そのモデルとその同期に関する実現の形式的仕様記述を行った., 1998年11月26日, 1998, 105, 87, 92
日本語, 情報処理学会論文誌. 数理モデル化と応用, 一般社団法人情報処理学会, 「情報処理学会論文誌 : 数理モデル化と応用」の編集にあたって, 城 和貴; 北 栄輔, 2006年08月15日, 47, 15, i
日本語, 情報処理学会研究報告, 情報処理学会, 小学生の奥行き知覚発達過程のモデル化—Modeling the Developmental Process of Depth Perception of Elementary Students—数理モデル化と問題解決(MPS) Vol.2009-MPS-74, 吉井 直子; 高田 雅美; 城 和貴, 2009年08月, 2009年度, 2, 1, 6
その他, 日本教育情報学会年会論文集, 幼児期の図形教育に関する研究, 吉井直子; 篠澤和恵; 高田雅美; 城和貴, 2019年, 35th
その他, 日本教育情報学会年会論文集, 幼児期における空間認識能力と図形教育, 吉井直子; 篠澤和恵; 高田雅美; 駒谷昇一; 城和貴, 2017年, 33rd