2024/09/25 更新

お知らせ

 

写真a

スエヒロ ダイキ
末廣 大貴
SUEHIRO DAIKI
所属
システム情報科学研究院 情報知能工学部門 准教授
工学部 電気情報工学科(併任)
システム情報科学府 情報理工学専攻(併任)
マス・フォア・イノベーション連係学府 (併任)
職名
准教授
連絡先
メールアドレス
電話番号
0928023574
プロフィール
統計的機械学習理論(理論的な汎化性能や計算量の保証)とその応用: - 学習問題の帰着 - ランキング学習 - マルチインスタンス学習 - (時)系列データ分析 オンライン意思決定問題とその応用: - オンライン物体認識 - 適応的パラメータチューニング コンピュータ将棋 ラーニングアナリティクス
外部リンク

学位

  • 博士(情報科学)

経歴

  • 横浜市立大学 データサイエンス学部 准教授

    2023年4月 - 2024年3月

      詳細を見る

  • 2014年4月~2016年4月 株式会社東芝 研究開発センター システム技術ラボラトリー

    2014年4月~2016年4月 株式会社東芝 研究開発センター システム技術ラボラトリー

研究テーマ・研究キーワード

  • 研究テーマ:機械学習

    研究キーワード:機械学習

    研究期間: 2024年

  • 研究テーマ:統計的学習, オンライン意思決定問題, 局所特徴分析, 時系列分析に関する理論ならびに応用研究

    研究キーワード:統計的学習, オンライン意思決定問題, 局所特徴分析, 時系列分析

    研究期間: 2017年4月

受賞

  • 電子情報通信学会九州支部連合大会講演奨励賞

    2022年9月   適応的データバランス調整~オンライン予測の理論に基づくアプローチ~

    斉藤優也, 内田誠一, 末廣大貴

     詳細を見る

  • Best Student Paper Award

    2022年5月   DAS2022   Revealing Reliable Signatures by Learning Top-Rank Pairs

    Xiaotong Ji, Yan Zheng, Daiki Suehiro, Seiichi Uchida

     詳細を見る

  • H29年度電子情報通信学会九州支部講演奨励賞受賞論文

    2017年9月   電子情報通信学会九州支部   Shapeletに基づいた文字認識

  • H29年度電子情報通信学会九州支部講演奨励賞受賞

    2017年9月   電気・情報関係学会九州支部連合  

  • Honorable Mention

    2011年11月   IBIS 2011   Honorable Mention

  • IBIS 2011 ポスター奨励賞 Honorable Mention

    2011年11月   第14回情報論的学習理論ワークショップ(IBIS2011)  

▼全件表示

論文

  • No regret sample selection with noisy labels 査読

    Song, H; Mitsuo, N; Uchida, S; Suehiro, D

    MACHINE LEARNING   113 ( 3 )   1163 - 1188   2024年3月   ISSN:0885-6125 eISSN:1573-0565

     詳細を見る

    担当区分:最終著者, 責任著者   掲載種別:研究論文(学術雑誌)   出版者・発行元:Machine Learning  

    Deep neural networks (DNNs) suffer from noisy-labeled data because of the risk of overfitting. To avoid the risk, in this paper, we propose a novel DNN training method with sample selection based on adaptive k-set selection, which selects k (< n, where n is the number of training samples) samples with a small noise-risk from the whole n noisy training samples at each epoch. It has the strong advantage of guaranteeing the performance of the selection theoretically. Roughly speaking, a regret, which is defined by the difference between the actual selection and the best selection, of the proposed method is theoretically bounded, even though the best selection is unknown until the end of all epochs. The experimental results on multiple noisy-labeled datasets demonstrate that our sample selection strategy works effectively in the DNN training; in fact, the proposed method achieved the best or the second-best performance among state-of-the-art methods, while requiring a significantly lower computational cost.

    DOI: 10.1007/s10994-023-06478-8

    Web of Science

    Scopus

    researchmap

  • COUNTING NETWORK FOR LEARNING FROM MAJORITY LABEL

    Shiku K., Matsuo S., Suehiro D., Bise R.

    ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings   7025 - 7029   2024年   ISSN:15206149 ISBN:9798350344851

     詳細を見る

    出版者・発行元:ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings  

    The paper proposes a novel problem in multi-class Multiple-Instance Learning (MIL) called Learning from the Majority Label (LML). In LML, the majority class of instances in a bag is assigned as the bag's label. LML aims to classify instances using bag-level majority classes. This problem is valuable in various applications. Existing MIL methods are unsuitable for LML due to aggregating confidences, which may lead to inconsistency between the bag-level label and the label obtained by counting the number of instances for each class. This may lead to incorrect instance-level classification. We propose a counting network trained to produce the bag-level majority labels estimated by counting the number of instances for each class. This led to the consistency of the majority class between the network outputs and one obtained by counting the number of instances. Experimental results show that our counting network outperforms conventional MIL methods on four datasets.

    DOI: 10.1109/ICASSP48485.2024.10448425

    Scopus

  • Paired contrastive feature for highly reliable offline signature verification 査読

    Ji, XT; Suehiro, D; Uchida, S

    PATTERN RECOGNITION   144   109816 - 109816   2023年12月   ISSN:0031-3203 eISSN:1873-5142

     詳細を見る

    掲載種別:研究論文(学術雑誌)   出版者・発行元:Pattern Recognition  

    Signature verification requires high reliability. Especially in the writer-independent scenario with the skilled-forgery-only condition, achieving high reliability is challenging but very important. In this paper, we propose to apply two machine learning frameworks, learning with rejection and top-rank learning, to this task because they can suppress ambiguous results and thus give only reliable verification results. Since those frameworks accept a single input, we transform a pair of genuine and query signatures into a single feature vector, called Paired Contrastive Feature (PCF). PCF internally represents similarity (or discrepancy) between the two paired signatures; thus, reliable machine learning frameworks can make reliable decisions using PCF. Through experiments on three public signature datasets in the offline skilled-forgery-only writer-independent scenario, we evaluate and validate the effectiveness and reliability of the proposed models by comparing their performance with a state-of-the-art model.

    DOI: 10.1016/j.patcog.2023.109816

    Web of Science

    Scopus

    researchmap

  • MixBag: Bag-Level Data Augmentation for Learning from Label Proportions 査読

    Takanori Asanomi, Shinnosuke Matsuo, Daiki Suehiro, Ryoma Bise

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023   16570 - 16579   2023年10月

     詳細を見る

  • Learning from Label Proportion with Online Pseudo-Label Decision by Regret Minimization 査読

    Matsuo S., Bise R., Uchida S., Suehiro D.

    ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings   2023年6月   ISSN:15206149

     詳細を見る

    担当区分:最終著者   掲載種別:研究論文(国際会議プロシーディングス)   出版者・発行元:ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings  

    This paper proposes a novel and efficient method for Learning from Label Proportions (LLP), whose goal is to train a classifier only by using the class label proportions of instance sets, called bags. We propose a novel LLP method based on an online pseudo-labeling method with regret minimization. As opposed to the previous LLP methods, the proposed method effectively works even if the bag sizes are large. We demonstrate the effectiveness of the proposed method using some benchmark datasets.

    DOI: 10.1109/icassp49357.2023.10097069

    Scopus

    researchmap

  • Boosting for Bounding the Worst-class Error.

    Yuya Saito, Shinnosuke Matsuo, Seiichi Uchida, Daiki Suehiro

    CoRR   abs/2310.14890   2023年

     詳細を見る

    掲載種別:研究論文(学術雑誌)  

    DOI: 10.48550/arXiv.2310.14890

    researchmap

  • Simplified and Unified Analysis of Various Learning Problems by Reduction to Multiple-Instance Learning 査読 国際誌

    Daiki Suehiro, Eiji Takimoto

    Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence, PMLR   180   1896 - 1906   2022年8月

     詳細を見る

    記述言語:その他   掲載種別:研究論文(国際会議プロシーディングス)  

  • Revealing Reliable Signatures by Learning Top-Rank Pairs 査読

    Ji, XT; Zheng, Y; Suehiro, D; Uchida, S

    DOCUMENT ANALYSIS SYSTEMS, DAS 2022   13237   323 - 337   2022年   ISSN:0302-9743 ISBN:978-3-031-06554-5 eISSN:1611-3349

     詳細を見る

    掲載種別:研究論文(国際会議プロシーディングス)   出版者・発行元:Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)  

    Signature verification, as a crucial practical documentation analysis task, has been continuously studied by researchers in machine learning and pattern recognition fields. In specific scenarios like confirming financial documents and legal instruments, ensuring the absolute reliability of signatures is of top priority. In this work, we proposed a new method to learn “top-rank pairs” for writer-independent offline signature verification tasks. By this scheme, it is possible to maximize the number of absolutely reliable signatures. More precisely, our method to learn top-rank pairs aims at pushing positive samples beyond negative samples, after pairing each of them with a genuine reference signature. In the experiment, BHSig-B and BHSig-H datasets are used for evaluation, on which the proposed model achieves overwhelming better pos@top (the ratio of absolute top positive samples to all of the positive samples) while showing encouraging performance on both Area Under the Curve (AUC) and accuracy.

    DOI: 10.1007/978-3-031-06555-2_22

    Web of Science

    Scopus

    researchmap

    その他リンク: https://dblp.uni-trier.de/db/conf/das/das2022.html#JiZSU22

  • Top-Rank Learning Robust to Outliers 査読 国際誌

    Yan Zheng, Daiki Suehiro, Seiichi Uchida

    The 28th International Conference on Neural Information Processing (ICONIP 2021)   2021年12月

     詳細を見る

    記述言語:英語   掲載種別:研究論文(国際会議プロシーディングス)  

  • Top-rank convolutional neural network and its application to medical image-based diagnosis.

    Yan Zheng, Yuchen Zheng, Daiki Suehiro, Seiichi Uchida

    Pattern Recognit.   120   108138 - 108138   2021年12月

     詳細を見る

    記述言語:その他   掲載種別:研究論文(学術雑誌)  

    DOI: 10.1016/j.patcog.2021.108138

  • Cell Detection from Imperfect Annotation by Pseudo Label Selection Using P-classification. 査読

    Kazuma Fujii, Daiki Suehiro, Kazuya Nishimura, Ryoma Bise

    International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI2021)   425 - 434   2021年10月

     詳細を見る

    記述言語:その他   掲載種別:研究論文(その他学術会議資料等)  

    DOI: 10.1007/978-3-030-87237-3_41

  • Theory and Algorithms for Shapelet-Based Multiple-Instance Learning 査読 国際誌

    Daiki Suehiro, Kohei Hatano, Eiji Takimoto, Shuji Yamamoto, Kenichi Bannai, Akiko Takeda

    Neural Computation   32 ( 8 )   1580 - 1613   2020年8月

     詳細を見る

    記述言語:英語   掲載種別:研究論文(学術雑誌)  

    DOI: 10.1162/neco_a_01297

  • Adaptive Aggregation of Arbitrary Online Trackers with a Regret Bound", Proceedings of the IEEE Winter Conference on Applications of Computer Vision 査読

    Heon Song, Daiki Suehiro, Seiichi Uchida

    Proceedings of the IEEE Winter Conference on Applications of Computer Vision   2020年3月

     詳細を見る

    記述言語:英語  

    Adaptive Aggregation of Arbitrary Online Trackers with a Regret Bound", Proceedings of the IEEE Winter Conference on Applications of Computer Vision

  • Adaptive aggregation of arbitrary online trackers with a regret bound

    Heon Song, Daiki Suehiro, Seiichi Uchida

    Proceedings - 2020 IEEE Winter Conference on Applications of Computer Vision, WACV 2020   670 - 678   2020年3月

     詳細を見る

    記述言語:その他   掲載種別:研究論文(その他学術会議資料等)  

    DOI: 10.1109/WACV45572.2020.9093613

  • Optimal Rejection Function Meets Character Recognition Tasks 査読

    Xiaotong Ji, Yuchen Zheng, Daiki Suehiro, Seiichi Uchida

    Proceedings of the 5th Asian Conference on Pattern Recognition   169 - 183   2019年11月

     詳細を見る

    記述言語:英語   掲載種別:研究論文(その他学術会議資料等)  

    Optimal Rejection Function Meets Character Recognition Tasks

    DOI: 10.1007/978-3-030-41299-9_14

  • Logo Design Analysis by Ranking. 査読

    Takuro Karamatsu, Daiki Suehiro, Seiichi Uchida

    1482 - 1487   2019年9月

     詳細を見る

    記述言語:その他   掲載種別:研究論文(その他学術会議資料等)  

    DOI: 10.1109/ICDAR.2019.00238

  • RankSVM for Offline Signature Verification. 査読

    Yan Zheng, Yuchen Zheng, Wataru Ohyama, Daiki Suehiro, Seiichi Uchida

    928 - 933   2019年9月

     詳細を見る

    記述言語:その他   掲載種別:研究論文(その他学術会議資料等)  

    DOI: 10.1109/ICDAR.2019.00153

  • Efficient reformulation of 1-norm ranking SVM 査読

    Daiki Suehiro, kohei hatano, Eiji Takimoto

    IEICE Transactions on Information and Systems   E101D ( 3 )   719 - 729   2018年3月

     詳細を見る

    記述言語:英語   掲載種別:研究論文(学術雑誌)  

    DOI: 10.1587/transinf.2017EDP7233

  • Face-to-Face Teaching Analytics: Extracting Teaching Activities from E-Book Logs via Time-Series Analysis 査読

    Daiki Suehiro, Yuta Taniguchi, Atsushi Shimada, Hiroaki Ogata

    Proceedings - IEEE 17th International Conference on Advanced Learning Technologies, ICALT 2017   267 - 268   2017年8月

     詳細を見る

    記述言語:英語   掲載種別:研究論文(その他学術会議資料等)  

    DOI: 10.1109/ICALT.2017.75

  • Revealing Hidden Impression Topics in Students' Journals Based on Nonnegative Matrix Factorization 査読

    Yuta Taniguchi, Daiki Suehiro, Atsushi Shimada, Hiroaki Ogata

    Proceedings - IEEE 17th International Conference on Advanced Learning Technologies, ICALT 2017   298 - 300   2017年8月

     詳細を見る

    記述言語:英語   掲載種別:研究論文(その他学術会議資料等)  

    DOI: 10.1109/ICALT.2017.113

  • M2B System: A Digital Learning Platform for Traditional Classrooms in University 査読

    Hiroaki Ogata, Yuta Taniguchi, Daiki Suehiro, Atsushi Shimada, Misato Oi, Fumiya Okubo, Masanori Yamada, Kentaro Kojima

    Practitioner Track Proceedings of the Seventh International Learning Analytics & Knowledge Conference   155 - 162   2017年3月

     詳細を見る

    記述言語:英語   掲載種別:研究論文(その他学術会議資料等)  

    M2B System: A Digital Learning Platform for Traditional Classrooms in University

  • Real-time learning analytics for C programming language courses 査読

    Xinyu Fu, Atsushi Shimada, Hiroaki Ogata, Yuta Taniguchi, Daiki Suehiro

    ACM International Conference Proceeding Series   280 - 288   2017年3月

     詳細を見る

    記述言語:英語   掲載種別:研究論文(その他学術会議資料等)  

    DOI: 10.1145/3027385.3027407

  • SVM による 2 部ランキング学習を用いたコンピュータ将棋における評価関数の学習 査読 国際誌

    末廣 大貴,畑埜 晃平,坂内 英夫,瀧本 英二,竹田 正幸

    電子情報通信学会論文誌 D, 情報・システム J97-D(3)   2014年3月

     詳細を見る

    記述言語:日本語   掲載種別:研究論文(学術雑誌)  

    リポジトリ公開URL: http://hdl.handle.net/2324/1505699

  • SVMによる2部ランキング学習を用いたコンピュータ将棋における評価関数の学習(情報・システム基礎) 査読

    末廣 大貴, 畑埜 晃平, 坂内 英夫, 瀧本 英二, 竹田 正幸

    電子情報通信学会論文誌. D, 情報・システム   97 ( 3 )   593 - 600   2014年3月

     詳細を見る

    記述言語:日本語   掲載種別:研究論文(学術雑誌)  

    近年,将棋の評価関数の設計においては,機械学習を応用してパラメータの自動調整を行う手法が主流となっている.ただし,評価項目(特徴)は作成者の棋力,感覚に基づいて用意されることが多く,これまで,複数の駒同士の関係など,複雑な特徴が数多く考案されてきた.本研究では,明示的に用意する特徴としては局面を表す基本的で単純なもののみとし,多項式カーネルとサポートベクターマシン(SVM)を用いて評価関数の学習を行う手法を提案する.多項式カーネルを用いることにより,単項式で表現できる特徴間のn項関係を,全て高次の特徴として利用することができる.また,評価関数の学習問題を,合法手後の局面を順位づける2部ランキングの問題と捉え,SVMを用いて学習を行う手法(ランキングSVM法)を提案する.対局や棋譜との一致率を調べる実験結果,及び駒組みの観察等から,ランキングSVM法の有効性を示す.

  • Online prediction under submodular constraints 査読

    Daiki Suehiro, Kohei Hatano, Shuji Kijima, Eiji Takimoto, Kiyohito Nagano

    Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)   7568   260 - 274   2012年10月

     詳細を見る

    記述言語:英語   掲載種別:研究論文(その他学術会議資料等)  

    DOI: 10.1007/978-3-642-34106-9_22

  • Approximate Reduction from AUC Maximization to 1-Norm Soft Margin Optimization 査読

    Daiki Suehiro, Kohei Hatano, Eiji Takimoto

    ALGORITHMIC LEARNING THEORY   6925   324 - 337   2011年10月

     詳細を見る

    記述言語:英語   掲載種別:研究論文(その他学術会議資料等)  

    DOI: 10.1007/978-3-642-24412-4_26

▼全件表示

講演・口頭発表等

  • 識別器の斟酌学習

    本田康祐, 内田誠一,末廣大貴

    2021年8月 

     詳細を見る

    開催年月日: 2021年8月

    記述言語:日本語   会議種別:口頭発表(一般)  

    開催地:オンライン   国名:日本国  

  • オンライン予測による画像分類器の識別率の制御

    本田康祐, 内田誠一, 末廣大貴

    電気・情報関係学会九州支部連合大会  2020年9月 

     詳細を見る

    開催年月日: 2021年5月

    記述言語:日本語  

    国名:日本国  

  • オンライントラッカの統合について

    ソン ホン, 末廣大貴, 内田誠一

    画像の認識・理解シンポジウム(MIRU2019)  2019年7月 

     詳細を見る

    開催年月日: 2021年5月

    記述言語:日本語  

    国名:日本国  

  • オンラインエキスパート選択問題としての適応的学習率調整

    満尾 成亮, 末廣 大貴, 内田 誠一

    画像の認識・理解シンポジウム(MIRU2019)  2019年7月 

     詳細を見る

    開催年月日: 2021年5月

    記述言語:日本語  

    国名:日本国  

  • 深層特徴を用いたリジェクション学習

    画像の認識・理解シンポジウム(MIRU2020)  2020年8月 

     詳細を見る

    開催年月日: 2021年5月

    記述言語:日本語  

    国名:日本国  

  • 任意のオンライントラッカの統合法

    画像の認識・理解シンポジウム(MIRU2020)  2020年8月 

     詳細を見る

    開催年月日: 2021年5月

    記述言語:日本語  

    国名:日本国  

  • GANを用いた局所パターン生成

    画像の認識・理解シンポジウム(MIRU2020)  2020年9月 

     詳細を見る

    開催年月日: 2021年5月

    記述言語:日本語  

    国名:日本国  

  • Adaptive aggregation of arbitrary online trackers with a regret bound

    Heon Song, Daiki Suehiro, Seiichi Uchida

    2020 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2020  2020年3月 

     詳細を見る

    開催年月日: 2020年3月

    記述言語:英語  

    国名:アメリカ合衆国  

  • Optimal Rejection Function Meets Character Recognition Tasks

    Xiaotong Ji, Yuchen Zheng, Daiki Suehiro, Seiichi Uchida

    5th Asian Conference on Pattern Recognition, ACPR 2019  2020年1月 

     詳細を見る

    開催年月日: 2019年11月

    記述言語:英語  

    国名:ニュージーランド  

  • Logo design analysis by ranking

    Takuro Karamatsu, Daiki Suehiro, Seiichi Uchida

    15th IAPR International Conference on Document Analysis and Recognition, ICDAR 2019  2019年9月 

     詳細を見る

    開催年月日: 2019年9月

    記述言語:英語  

    国名:オーストラリア連邦  

  • RankSVM for offline signature verification

    Yan Zheng, Yuchen Zheng, Wataru Ohyama, Daiki Suehiro, Seiichi Uchida

    15th IAPR International Conference on Document Analysis and Recognition, ICDAR 2019  2019年9月 

     詳細を見る

    開催年月日: 2019年9月

    記述言語:英語  

    国名:オーストラリア連邦  

  • Online People-flow Prediction

    Heon Song, Daiki Suehiro, Seiichi Uchida

    2018年9月 

     詳細を見る

    開催年月日: 2019年6月

    記述言語:英語  

    国名:日本国  

  • オンラインエキスパート統合アルゴリズムに基づく異常検知

    満尾成亮, 末廣大貴, 内田誠一

    電気・情報関係学会九州支部連合大会  2018年9月 

     詳細を見る

    開催年月日: 2019年6月

    記述言語:日本語  

    開催地:大分   国名:日本国  

  • Face-to-Face Teaching Analytics Extracting Teaching Activities from E-Book Logs via Time-Series Analysis

    Daiki Suehiro, Yuta Taniguchi, Atsushi Shimada

    17th IEEE International Conference on Advanced Learning Technologies, ICALT 2017  2017年8月 

     詳細を見る

    開催年月日: 2017年7月

    記述言語:英語  

    国名:その他  

    リポジトリ公開URL: http://hdl.handle.net/2324/4068613

  • Revealing Hidden Impression Topics in Students' Journals Based on Nonnegative Matrix Factorization

    Yuta Taniguchi, Daiki Suehiro, Atsushi Shimada

    17th IEEE International Conference on Advanced Learning Technologies, ICALT 2017  2017年8月 

     詳細を見る

    開催年月日: 2017年7月

    記述言語:英語  

    国名:その他  

  • Real-time learning analytics for C programming language courses

    Xinyu Fu, Atsushi Shimada, Hiroaki Ogata, Yuta Taniguchi, Daiki Suehiro

    7th International Conference on Learning Analytics and Knowledge, LAK 2017  2017年3月 

     詳細を見る

    開催年月日: 2017年3月

    記述言語:英語  

    国名:カナダ  

  • Online prediction under submodular constraints

    Daiki Suehiro, kohei hatano, Shuji Kijima, Eiji Takimoto, Kiyohito Nagano

    23rd International Conference on Algorithmic Learning Theory, ALT 2012  2012年10月 

     詳細を見る

    開催年月日: 2012年10月

    記述言語:英語  

    国名:フランス共和国  

    リポジトリ公開URL: http://hdl.handle.net/2324/1932327

  • Approximate reduction from AUC maximization to 1-norm soft margin optimization

    Daiki Suehiro, kohei hatano, Eiji Takimoto

    22nd International Conference on Algorithmic Learning Theory, ALT 2011  2011年10月 

     詳細を見る

    開催年月日: 2011年10月

    記述言語:英語  

    国名:フィンランド共和国  

    リポジトリ公開URL: http://hdl.handle.net/2324/1546621

  • 局所パターン生成の検討

    Chean Fei Shee, 末廣大貴, 内田誠一

    電気・情報関係学会九州支部連合大会  2019年9月 

     詳細を見る

    開催年月日: - 2021年5月

    記述言語:英語  

    国名:日本国  

  • Learning theory and algorithms for shapelets and other local features 国際会議

    Daiki Suehiro, Kohei Hatano, Eiji Takimoto, Shuji Yamamoto, Kenichi Bannai, Akiko Takeda

    NIPS Time Series Workshop 2017,  2017年12月 

     詳細を見る

    記述言語:英語   会議種別:シンポジウム・ワークショップ パネル(公募)  

    国名:アメリカ合衆国  

  • 投手の打ちづらさとは何か ~ 機械学習に基づく投球印象解析 ~

    角淳之介, 末廣大貴, 加藤貴昭, 内田誠一

    スポーツ情報処理時限研究会  2018年12月 

     詳細を見る

    記述言語:日本語  

    開催地:東京   国名:日本国  

  • 弱教師学習問題における最適局所特徴抽出および樹状突起スパイン検出への応用

    八尋俊希,末廣大貴,本館利佳,鈴木利治,内田誠一

    医用画像研究会  2019年1月 

     詳細を見る

    記述言語:日本語  

    開催地:沖縄   国名:日本国  

  • Shapelet-based Multiple-Instance Learning

    Daiki Suehiro, Kohei Hatano, Eiji Takimoto, Shuji Yamamoto, Kenichi Bannai, Akiko Takeda

    2019年3月 

     詳細を見る

    記述言語:日本語  

    国名:日本国  

▼全件表示

MISC

  • オンライン予測による画像分類器の識別率の制御

    本田康祐, 内田誠一, 末廣大貴

    電気・情報関係学会九州支部連合大会講演論文集(CD-ROM)   2020年9月

     詳細を見る

    記述言語:その他  

  • Reduction Scheme for Empirical Risk Minimization and Its Applications to Multiple-Instance Learning

    Daiki Suehiro, Eiji Takimoto

    2019年11月

     詳細を見る

    記述言語:その他  

  • 弱教師学習問題における最適局所特徴抽出および樹状突起スパイン検出への応用

    八尋俊希, 末廣大貴, 末廣大貴, 本館利佳, 鈴木利治, 内田誠一

    電子情報通信学会技術研究報告   2019年1月

     詳細を見る

    記述言語:日本語  

  • 投手の打ちづらさとは何か ~ 機械学習に基づく投球印象解析 ~

    角淳之介, 末廣大貴, 加藤貴昭, 内田誠一

    映像情報処理学会メディア工学研究会(ME) 2018-114   2018年12月

     詳細を見る

    記述言語:日本語  

  • オンラインエキスパート統合アルゴリズムに基づく異常検知

    満尾成亮, 末廣大貴, 内田誠一

    電気・情報関係学会九州支部連合大会講演論文集(CD-ROM)   2018年9月

     詳細を見る

    記述言語:日本語  

  • Online People-flow Prediction

    Heon Song, Daiki Suehiro, Seiichi Uchida

    2018年9月

     詳細を見る

    記述言語:英語  

    Online People-flow Prediction

  • Learning theory and algorithms for shapelets and other local features 査読

    Daiki Suehiro, Kohei Hatano, Eiji Takimoto, Shuji Yamamoto, Kenichi Bannai, Akiko Takeda

    NIPS Time Series Workshop 2017   2017年12月

     詳細を見る

    記述言語:その他  

    Learning theory and algorithms for shapelets and other local features

  • Shapeletに基づいた文字認識

    八尋俊希, 末廣大貴, 内田誠一

    電気・情報関係学会九州支部連合大会講演論文集(CD-ROM)   2017年9月

     詳細を見る

    記述言語:日本語  

  • Boosting the kernelized shapelets: Theory and algorithms for local features

    Daiki Suehiro, Kohei Hatano, Eiji Takimoto, Shuji Yamamoto, Kenichi Bannai, Akiko Takeda

    2017年9月

     詳細を見る

    記述言語:その他  

    Boosting the kernelized shapelets: Theory and algorithms for local features
    We consider binary classification problems using local features of objects.<br />
    One of motivating applications is time-series classification, where features<br />
    reflecting some local closeness measure between a time series and a pattern<br />
    sequence called shapelet are useful. Despite the empirical success of such<br />
    approaches using local features, the generalization ability of resulting<br />
    hypotheses is not fully understood and previous work relies on a bunch of<br />
    heuristics. In this paper, we formulate a class of hypotheses using local<br />
    features, where the richness of features is controlled by kernels. We derive<br />
    generalization bounds of sparse ensembles over the class which is exponentially<br />
    better than a standard analysis in terms of the number of possible local<br />
    features. The resulting optimization problem is well suited to the boosting<br />
    approach and the weak learning problem is formulated as a DC program, for which<br />
    practical algorithms exist. In preliminary experiments on time-series data<br />
    sets, our method achieves competitive accuracy with the state-of-the-art<br />
    algorithms with small parameter-tuning cost.

  • Time Series Classification Based on Random Shapelets 査読

    Daiki Suehiro, Kengo Kuwahara, Kohei Hatano, Eiji Takimoto

    NIPS Time Series Workshop 2016   2016年12月

     詳細を見る

    記述言語:英語  

    Time Series Classification Based on Random Shapelets

  • 教育データのオープン化に向けて (パターン認識・メディア理解)

    末廣 大貴, 毛利 考佑, 谷口 雄太, 大久保 文哉, 島田 敬士, 緒方 広明

    電子情報通信学会技術研究報告 = IEICE technical report : 信学技報   2016年10月

     詳細を見る

    記述言語:日本語  

  • ランキングSVMの近似に基づく効率的なAUC最大化(第15回情報論的学習理論ワークショップ)

    末廣 大貴, 畑埜 晃平, 瀧本 英二

    電子情報通信学会技術研究報告. IBISML, 情報論的学習理論と機械学習   2012年10月

     詳細を見る

    記述言語:英語  

    The formulation of Ranking SVMs is popular for maximizing AUC scores. More precisely, the formulation is given as a hard/soft margin optimization over pn pairs of p positive and n negative instances. Directly solving the problem is impractical since we have to deal with a sample of size pn, which is quadratically larger than the original sample size p+n. In this paper, we propose (approximate) reduction methods from the hard/soft margin optimization over pn pairs to variants of hard/soft margin optimization over p+n instances. The resulting classifiers of our methods are guaranteed to have a certain amount of margin over pn pairs.

  • 劣モジュラ制約下におけるオンライン予測

    末廣 大貴, 畑埜 晃平, 来嶋 秀治, 瀧本 英二, 永野 清仁

    電子情報通信学会技術研究報告. IBISML, 情報論的学習理論と機械学習 = IEICE technical report. IBISML, Information-based induction sciences and machine learning   2012年6月

     詳細を見る

    記述言語:英語  

    本論文では離散構造のオンライン予測問題を考える.ただし,各離散構造は劣モジュラ基多面体の頂点で表現できると仮定する.例として,全域木や,順列のオンライン予測問題がある.一般的に基多面体の端点は指数個存在するが,本論文では多項式時間オンライン予測アルゴリズムを提案し,リグレットの評価を与える.特に,劣モジュラ関数が入力のサイズのみに依存するとき,提案アルゴリズムの1つはO(n^2)時間で動作する.

  • Approximate Reduction from AUC Maximization to 1-norm Soft Margin Optimization (情報論的学習理論と機械学習)

    電子情報通信学会技術研究報告 : 信学技報   2011年11月

     詳細を見る

    記述言語:英語  

    Finding linear classifiers that maximize AUC scores is important in ranking research. This is naturally formulated as a 1-norm hard/soft margin optimization problem over pn pairs of p positive and n negative instances. However, directly solving the optimization problems is impractical since the problem size (pn) is quadratically larger than the given sample size (p+n). In this paper, we give (approximate) reductions from the problems to hard/soft margin optimization problems of linear size. First, for the hard margin case, we show that the problem is reduced to a hard margin optimization problem over p+n instances in which the bias constant term is to be optimized. Then, for the soft margin case, we show that the problem is approximately reduced to a soft margin optimization problem over p+n instances for which the resulting linear classifier is guaranteed to have a certain margin over pairs.

  • カーネル法を用いたコンピュータ将棋の評価関数の学習 査読

    末廣 大貴, 畑埜 晃平, 坂内 英夫, 瀧本 英二, 竹田 正幸

    ゲームプログラミングワークショップ2010論文集   2010年11月

     詳細を見る

    記述言語:日本語  

    近年,コンピュータ将棋における評価関数は,機械学習を応用したパラメータの自動調整を行う手法が主流となっている.ただし,評価項目(特徴)は,作成者の考え,感覚に基づいて用意されることがほとんどである.本論文では,プロ棋士の棋譜を学習サンプルとして,カーネル法とサポートベクトルマシンを用いて学習を行う手法を提案する.カーネル法を用いることにより,作成者があらかじめ複雑な特徴を用意せずとも,局面を表現する単純な特徴のみから,特徴間のn項関係などのより高次な特徴のが暗に生成され,その特徴空間で学習が行われる.複数の駒の位置関係の考慮が不可欠である囲いの評価実験を行い,カーネル法が有用性を示す結果を得た.また,本手法により得られた評価関数は,定跡などの明示的な知識を導入することなしに得られたにもかかわらず,特に序盤において,人間らしい局面評価を行うことを示す.Recently, automatic optimization of parameters by applying machine learning methods has become a mainstream approach for developing good evaluation functions in shogi. However, the features used in the evaluation functions are prepared by the developer, depending heavily on his/her knowledge and intuition. In this paper, we propose a method for learning evaluation functions from game records of professional players, using kernel methods and Support Vector Machines (SVMs). By using kernels, higher dimensional features such as n-ary relations between simple features can be considered implicitly, and various complex features can be considered without preparation. We apply our method on castle positions, which require consideration of relative positions of pieces, and show that the evaluation functions learned using kernels give better results. We also show that even without knowledge of standard moves, we were able to obtain human-like evaluation functions, especially in the opening.

  • SVMによるバイパータイトランキング学習を用いたコンピュータ将棋における評価関数の学習(IBIS2010(情報論的学習理論ワークショップ))

    末廣 大貴, 畑埜 晃平, 坂内 英夫, 瀧本 英二, 竹田 正幸

    電子情報通信学会技術研究報告. IBISML, 情報論的学習理論と機械学習   2010年10月

     詳細を見る

    記述言語:日本語  

    近年,将棋の評価関数の設計においては,機械学習を応用してパラメータの自動調整を行う手法が主流となっている.ただし,評価項目(特徴)は作成者の棋力,感覚に基づいて用意されることが多く,これまで,複数の駒同士の関係など,複雑な特徴が数多く考案されてきた.本研究では,明示的に用意する特徴としては局面を表す基本的で単純なもののみとし,多項式カーネルとサポートベクターマシン(SVM)を用いて評価関数の学習を行う手法を提案する.多項式カーネルを用いることにより,単項式で表現できる特徴間のn項関係を,すべて高次の特徴として利用することができる・また,評価関数の学習問題を,合法手後の局面を順位づけるバイパータイトランキングの問題と捉え,SVMを用いて学習を行う手法(ランキングSVM法)を提案する.対局や棋譜との一致率を調べる実験結果,および駒組みの観察等から,ランキングSVM法の有効性を示す.

▼全件表示

産業財産権

特許権   出願件数: 1件   登録件数: 1件
実用新案権   出願件数: 0件   登録件数: 0件
意匠権   出願件数: 0件   登録件数: 0件
商標権   出願件数: 0件   登録件数: 0件

所属学協会

  • 電子情報通信学会(情報・システムソサイエティ)

  • 電子情報通信学会

委員歴

  • 電子情報通信学会 システム数理と応用研究会(MSS研究会) 回路とシステムワークショップ D(離散システム理論)分科会   運営委員   国内

    2020年4月 - 現在   

  • 電子情報通信学会 システム数理と応用研究会(MSS研究会) 回路とシステムワークショップ D(離散システム理論)分科会   企画委員   国内

    2020年4月 - 現在   

  • ACML2019 Program Committee member  

    2019年 - 現在   

      詳細を見る

    団体区分:学協会

    researchmap

  • AAAI2019 Program Committee member  

    2018年 - 現在   

      詳細を見る

    団体区分:学協会

    researchmap

  • ACML2018 Program Committee member  

    2018年 - 現在   

      詳細を見る

    団体区分:学協会

    researchmap

  • 電子情報通信学会 システム数理と応用研究会(MSS研究会) 回路とシステムワークショップ D(離散システム理論)分科会   実行委員  

    2017年7月 - 現在   

      詳細を見る

    団体区分:学協会

    researchmap

  • 電子情報通信学会 システム数理と応用研究会(MSS研究会) 回路とシステムワークショップ D(離散システム理論)分科会   運営委員   国内

    2017年7月 - 2021年3月   

  • 電子情報通信学会 システム数理と応用研究会(MSS研究会) 回路とシステムワークショップ D(離散システム理論)分科会   世話人   国内

    2017年7月 - 2021年3月   

▼全件表示

学術貢献活動

  • 実行委員

    ( オンライン ) 2021年11月

     詳細を見る

    種別:大会・シンポジウム等 

  • 学術論文等の審査

    役割:査読

    2021年

     詳細を見る

    種別:査読等 

    外国語雑誌 査読論文数:1

  • PC member 国際学術貢献

    The 12th Asian Conference on Machine Learning  ( Online ) 2020年11月

     詳細を見る

    種別:大会・シンポジウム等 

  • 学術論文等の審査

    役割:査読

    2020年

     詳細を見る

    種別:査読等 

    外国語雑誌 査読論文数:1

    日本語雑誌 査読論文数:0

    国際会議録 査読論文数:6

    国内会議録 査読論文数:1

  • PC member 国際学術貢献

    The 11th Asian Conference on Machine Learning  ( Japan ) 2019年11月

     詳細を見る

    種別:大会・シンポジウム等 

  • PC member 国際学術貢献

    The Thirty-Third AAAI Conference on Artificial Intelligence (AAAI-19)  ( Honolulu, Hawaii UnitedStatesofAmerica ) 2019年1月 - 2019年2月

     詳細を見る

    種別:大会・シンポジウム等 

  • 学術論文等の審査

    役割:査読

    2019年

     詳細を見る

    種別:査読等 

    外国語雑誌 査読論文数:1

    国際会議録 査読論文数:3

    国内会議録 査読論文数:2

▼全件表示

共同研究・競争的資金等の研究課題

  • データサンプリングを前提とした機械学習の包括的枠組み

    研究課題/領域番号:24K03002  2024年4月 - 2029年3月

    科学研究費助成事業  基盤研究(B)

    末廣 大貴, 備瀬 竜馬

      詳細を見る

    資金種別:科研費

    本研究では学習問題における「学習目的」(分類精度,適合率,回帰精度の最大化など)と「教師データ条件」(教師ありデータ,半教師ありデータなど)に着目し,汎用化に取り組む. 具体的には,以下の3つを明らかにする.
    ①多様な学習目的に対応する汎用データサンプリングの枠組み
    ②多様な教師データ条件に対応する汎用データサンプリングの枠組み
    ③複数の教師データ条件に対応する汎用データサンプリングの枠組み

    CiNii Research

  • オンライン予測理論に基づくデータサンプリング問題への統合的アプローチ

    2021年4月 - 2024年3月

      詳細を見る

    担当区分:研究代表者 

    本研究の目的は,機械学習分野におけるデータサンプリング問題に対し,統合的な定式化と理論解析の枠組みを与えることにある.データサンプリングは,全てのサンプルを学習に用いるのではなく,可能な限り「望ましいデータ」のみをサンプリングするタスクのことで,多くのドメインで幅広く考えられているタスクである.ノイズデータを回避してモデルの頑健性を保ったり,重要なデータを選んで学習スピードを加速させたりと,データサンプリングには様々なタスクが存在する.しかし,ドメイン,タスクの細かい特性に応じたアドホックな定式化や手法が多く,汎用性や理論解析に関する議論が欠如している.本研究では,ドメイン,タスク依存の現状を打破するため,データサンプリング問題について,統合的な枠組みの開発,理論性能保証,実応用の開拓を行う.

  • オンライン予測理論に基づくデータサンプリング問題への統合的アプローチ

    研究課題/領域番号:21K12032  2021年 - 2023年

    日本学術振興会  科学研究費助成事業  基盤研究(C)

      詳細を見る

    資金種別:科研費

  • オンライン予測理論に基づくデータサンプリング問題への統合的アプローチ

    2021年 - 2023年

    日本学術振興会  科学研究費助成事業  基盤研究(C)

      詳細を見る

    担当区分:研究代表者  資金種別:科研費

  • オンライン予測理論に基づくデータサンプリング問題への統合的アプローチ

    研究課題/領域番号:21K12032  2021年 - 2023年

    日本学術振興会  科学研究費助成事業  基盤研究(C)

      詳細を見る

    担当区分:研究代表者  資金種別:科研費

  • 学習問題の統合的帰着

    2020年11月 - 2023年3月

    国立研究開発法人科学技術振興機構(日本) 

      詳細を見る

    担当区分:研究代表者 

    機械学習の分野では,様々なドメインで学習問題が提案,展開されている.理論的な解析はドメインごとに個別に行われていることが多く,理論解析が十分に行われていない問題も多々存在する.本研究では,様々な学習問題を別の学習問題へ「まとめて」帰着することにより,ドメインを超えた統合的な理論解析を目指す.まず,マルチインスタンス学習と呼ばれる学習問題への統合的帰着を研究の足がかりとする.また,統合的帰着手法の適用範囲の拡大し,帰着手法の一般形を究明する.

  • 生命科学特有の付加データ及びドメイン知識に着目した弱教師学習手法の開発

    2020年4月 - 2023年3月

      詳細を見る

    担当区分:研究分担者 

  • 生命科学特有の付加データ及びドメイン知識に着目した弱教師学習手法の開発

    2020年 - 2022年

    日本学術振興会  科学研究費助成事業  基盤研究(B)

      詳細を見る

    担当区分:研究分担者  資金種別:科研費

  • 学習問題の統合的帰着

    2020年 - 2022年

    日本学術振興会  科学研究費助成事業  基盤研究(C)

      詳細を見る

    担当区分:研究代表者  資金種別:科研費

  • 生命科学特有の付加データ及びドメイン知識に着目した弱教師学習手法の開発

    研究課題/領域番号:20H04211  2020年 - 2022年

    日本学術振興会  科学研究費助成事業  基盤研究(C)

    備瀬 竜馬, 末廣 大貴, 渡邉 和秀, 吉澤 明彦

      詳細を見る

    資金種別:科研費

    本研究課題では,生命科学分野でのLong-tailな研究課題の現場における深層学習の実活用の大きなボトルネックの一つとなっている教師データ作成に着目して,生命科学特有の簡易に取得可能な弱教師を活かすことが可能な問題設定及び手法開発を新たに行い,様々な課題で簡易に適用できるフレームワークを構築することで,実活用拡大につなげる.

    CiNii Research

  • 「学習問題の統合的帰着」

    2020年 - 2022年

    戦略的創造研究推進事業 (文部科学省)

      詳細を見る

    担当区分:研究代表者  資金種別:受託研究

  • 科研費(若手):「局所パターン学習とは何か:統一的定式化と理論性能保証」

    2018年4月 - 2020年3月

      詳細を見る

    担当区分:研究代表者 

  • 局所パターン学習とは何か:統一的定式化と理論性能保証

    研究課題/領域番号:18K18001  2018年 - 2019年

    日本学術振興会  科学研究費助成事業  若手研究

      詳細を見る

    担当区分:研究代表者  資金種別:科研費

  • 局所パターン学習とは何か:統一的定式化と理論性能保証

    研究課題/領域番号:18K18001  2018年 - 2019年

    日本学術振興会  科学研究費助成事業  基盤研究(C)

      詳細を見る

    資金種別:科研費

▼全件表示

教育活動概要

  • ・内田・備瀬研究室における学生指導補助
    ・大学院システム情報科学府 「データサイエンス演習第一」, 「データサイエンス演習第二」
    ・大学院特別講義 「サポートベクトルマシンと機械学習理論」
    ・基幹教育「サイバーセキュリティ基礎論」

担当授業科目

  • データサイエンス発展ⅠI

    2021年4月 - 2021年6月   春学期

  • サイバーセキュリティ基礎論

    2021年4月 - 2021年6月   春学期

  • データサイエンス実践Ⅰ

    2021年4月 - 2021年6月   春学期

  • データサイエンス実践ⅠI

    2021年4月 - 2021年6月   春学期

  • データサイエンス実践ⅠII

    2021年4月 - 2021年6月   春学期

  • データサイエンス実践ⅠV

    2021年4月 - 2021年6月   春学期

  • データサイエンス発展Ⅰ

    2021年4月 - 2021年6月   春学期

  • データサイエンス演習第一

    2020年10月 - 2021年3月   後期

  • データサイエンス演習第二

    2020年10月 - 2021年3月   後期

  • サイバーセキュリティ基礎論

    2020年4月 - 2020年6月   春学期

  • SVMで学ぶ機械学習理論

    2019年10月 - 2020年3月   後期

  • データサイエンス演習第一

    2019年4月 - 2019年9月   前期

  • データサイエンス演習第二

    2019年4月 - 2019年9月   前期

  • サイバーセキュリティ基礎論

    2019年4月 - 2019年6月   春学期

  • サポートベクトルマシンと機械学習理論

    2018年10月 - 2019年3月   後期

  • データサイエンス演習第二

    2018年4月 - 2018年9月   前期

  • データサイエンス演習第一

    2018年4月 - 2018年9月   前期

  • サポートベクトルマシンと機械学習理論

    2017年10月 - 2018年3月   後期

  • データサイエンス演習第一

    2017年4月 - 2017年9月   前期

  • データサイエンス演習第二

    2017年4月 - 2017年9月   前期

▼全件表示

FD参加状況

  • 2022年1月   役割:参加   名称:【シス情FD】シス情関連の科学技術に対する国の政策動向(に関する私見)

    主催組織:部局

  • 2021年9月   役割:参加   名称:博士後期課程の充足率向上に向けて

    主催組織:部局

  • 2021年7月   役割:参加   名称:若手教員による研究紹介 及び 科研取得のポイント、その他について ②

    主催組織:部局

  • 2020年9月   役割:参加   名称:電気情報工学科総合型選抜(AO入試)について

    主催組織:部局

  • 2019年10月   役割:参加   名称:電子ジャーナルの現状と今後の動向に関する説明会

    主催組織:部局

  • 2019年6月   役割:参加   名称:8大学情報系研究科長会議の報告

    主催組織:部局

  • 2017年10月   役割:参加   名称:いよいよスタートした電気情報工学科国際コース

    主催組織:部局

▼全件表示

他大学・他機関等の客員・兼任・非常勤講師等

  • 2020年  理化学研究所 革新知能統合研究センター 計算論的学習理論チーム  区分:客員教員  国内外の区分:国内 

  • 2019年  理化学研究所 革新知能統合研究センター 計算論的学習理論チーム  区分:客員教員  国内外の区分:国内 

  • 2018年  理化学研究所 革新知能統合研究センター 計算論的学習理論チーム  区分:客員教員  国内外の区分:国内 

その他教育活動及び特記事項

  • 2018年  その他特記事項  システム開発技術カレッジ公開講座データサイエンス概論(Python入門・演習),公益財団法人福岡産業・科学技術振興財団(ふくおかIST),システム技術カレッジ

     詳細を見る

    システム開発技術カレッジ公開講座データサイエンス概論(Python入門・演習),公益財団法人福岡産業・科学技術振興財団(ふくおかIST),システム技術カレッジ

社会貢献活動

  • システム開発技術カレッジ公開講座 データサイエンス概論(Python入門・演習)

    公益財団法人 福岡産業・科学技術振興財団(ふくおかアイスト)  システム開発技術カレッジ  2018年12月

     詳細を見る

    対象: 社会人・一般, 学術団体, 企業, 市民団体, 行政機関

    種別:講演会

学内運営に関わる各種委員・役職等

  • 2020年10月 - 現在   その他 九州大学大学院システム情報科学研究院ADS育成室(Advanced Data Scientist Training Unit)協力研究員