Authorship：Last author   Publishing type：Research paper (international conference proceedings)   Publisher：IEEE  

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

Publishing type：Research paper (scientific journal)  

DOI： 10.48550/arXiv.2310.14890

researchmap

Language：Others   Publishing type：Research paper (international conference proceedings)  

Publishing type：Research paper (international conference proceedings)   Publisher：Springer  

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

Other Link： https://dblp.uni-trier.de/db/conf/das/das2022.html#JiZSU22

Language：English   Publishing type：Research paper (international conference proceedings)  

Language：Others   Publishing type：Research paper (scientific journal)  

Top-rank learning identifies a real-valued ranking function that will provide more absolute top samples. These are highly reliable positive samples that are ranked higher than the highest-ranked negative samples. Therefore, top-rank learning is useful for tasks that require reliable decisions. Additionally, it inherits the merits of the ranking functions, such as robustness to the unbalanced condition. However, conventional top-rank learning tasks are formulated as linear or kernel-based problems and are thus limited in coping with complicated tasks. In this study, we propose a Top-rank convolutional neural network (TopRank CNN) to realize top-rank learning with representation learning for complicated tasks. Given that the original objective function of top-rank learning suffers from overfitting, we employ the p-norm relaxation of the original loss function in the proposed method. We prove the usefulness of TopRank CNN experimentally with medical diagnosis tasks that require reliable decisions and robustness to the unbalanced condition.

DOI： 10.1016/j.patcog.2021.108138

Language：Others   Publishing type：Research paper (other academic)  

DOI： 10.1007/978-3-030-87237-3_41

Language：English   Publishing type：Research paper (scientific journal)  

We propose a new formulation of multiple-instance learning (MIL), in which a unit of data consists of a set of instances called a bag. The goal is to find a good classifier of bags based on the similarity with a "shapelet" (or pattern), where the similarity of a bag with a shapelet is the maximum similarity of instances in the bag. In previous work, some of the training instances have been chosen as shapelets with no theoretical justification. In our formulation, we use all possible, and thus infinitely many, shapelets, resulting in a richer class of classifiers. We show that the formulation is tractable, that is, it can be reduced through linear programming boosting (LPBoost) to difference of convex (DC) programs of finite (actually polynomial) size. Our theoretical result also gives justification to the heuristics of some previous work. The time complexity of the proposed algorithm highly depends on the size of the set of all instances in the training sample. To apply to the data containing a large number of instances, we also propose a heuristic option of the algorithm without the loss of the theoretical guarantee. Our empirical study demonstrates that our algorithm uniformly works for shapelet learning tasks on time-series classification and various MIL tasks with comparable accuracy to the existing methods. Moreover, we show that the proposed heuristics allow us to achieve the result in reasonable computational time.

DOI： 10.1162/neco_a_01297

Language：Others   Publishing type：Research paper (other academic)  

We propose an online visual-object tracking method that is robust even in an adversarial environment, where various disturbances may occur on the target appearance, etc. The proposed method is based on a delayed-Hedge algorithm for aggregating multiple arbitrary online trackers with adaptive weights. The robustness in the tracking performance is guaranteed theoretically in term of "regret" by the property of the delayed-Hedge algorithm. Roughly speaking, the proposed method can achieve a similar tracking performance as the best one among all the trackers to be aggregated in an adversarial environment. The experimental study on various tracking tasks shows that the proposed method could achieve state-of-the-art performance by aggregating various online trackers.

DOI： 10.1109/WACV45572.2020.9093613

Language：English  

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

Language：English   Publishing type：Research paper (other academic)  

Optimal Rejection Function Meets Character Recognition Tasks

DOI： 10.1007/978-3-030-41299-9_14

Language：Others   Publishing type：Research paper (other academic)  

DOI： 10.1109/ICDAR.2019.00238

Language：Others   Publishing type：Research paper (other academic)  

DOI： 10.1109/ICDAR.2019.00153

Language：English   Publishing type：Research paper (scientific journal)  

Finding linear functions that maximize AUC scores is important in ranking research. A typical approach to the ranking problem is to reduce it to a binary classification problem over a new instance space, consisting of all pairs of positive and negative instances. Specifically, this approach is formulated as hard or soft margin optimization problems over pn pairs of p positive and n negative instances. Solving the optimization problems directly 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 reformulate the ranking problem as variants of hard and soft margin optimization problems over p+n instances. The resulting classifiers of our methods are guaranteed to have a certain amount of AUC scores.

DOI： 10.1587/transinf.2017EDP7233

Language：English   Publishing type：Research paper (other academic)  

Students' reflective writings are useful not only for students themselves but also teachers. It is important for teachers to know which concepts were understood well by students and which concepts were not, to continuously improve their classes. However, it is difficult for teachers to thoroughly read the journals of more than one hundred students. In this paper, we propose a novel method to extract common topics and students' common impressions against them from students' journals. Weekly keywords are discovered from journals by scoring noun words with a measure based on TF-IDF term weighting scheme, and then we analyze co-occurrence relationships between extracted keywords and adjectives. We employs nonnegative matrix factorization, one of the topic modeling techniques, to discover the hidden impression topics from the co-occurrence relationships. As a case study, we applied our method on students' journals of the course 'Information Science' held in our university. Our experimental results show that conceptual keywords are successfully extracted, and four significant impression topics are identified. We conclude that our analysis method can be used to collectively understand the impressions of students from journal texts.

DOI： 10.1109/ICALT.2017.113

Language：English   Publishing type：Research paper (other academic)  

To discover teaching knowledge efficiently, we must extract the various teaching activities from educational data. In this paper, through the use of e-book logs and techniques of time-series analysis, we describe a method of practicing teaching analytics in face-to-face classes, one which enable us to extract the teaching activity efficiently and accurately.

DOI： 10.1109/ICALT.2017.75

Language：English   Publishing type：Research paper (other academic)  

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

Language：English   Publishing type：Research paper (other academic)  

Many universities choose the C programming language (C) as the first one they teach their students, early on in their program. However, students often consider programming courses difficult, and these courses often have among the highest dropout rates of computer science courses offered. It is therefore critical to provide more effective instruction to help students understand the syntax of C and prevent them losing interest in programming. In addition, homework and paper-based exams are still the main assessment methods in the majority of classrooms. It is difficult for teachers to grasp students' learning situation due to the large amount of evaluation work. To facilitate teaching and learning of C, in this article we propose a system-LAPLE (Learning Analytics in Programming Language Education)-that provides a learning dashboard to capture the behavior of students in the classroom and identify the different difficulties faced by different students looking at different knowledge. With LAPLE, teachers may better grasp students' learning situation in real time and better improve educational materials using analysis results. For their part, novice undergraduate programmers may use LAPLE to locate syntax errors in C and get recommendations from educational materials on how to fix them.

DOI： 10.1145/3027385.3027407

Language：Japanese   Publishing type：Research paper (scientific journal)  

Repository Public URL： http://hdl.handle.net/2324/1505699

Language：Japanese   Publishing type：Research paper (scientific journal)  

Learning Evaluation Functions for Shogi Using SVM-Based Bipartite Ranking Learning
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.

Language：English   Publishing type：Research paper (other academic)  

We consider an online prediction problem of combinatorial concepts where each combinatorial concept is represented as a vertex of a polyhedron described by a submodular function (base polyhedron). In general, there are exponentially many vertices in the base polyhedron. We propose polynomial time algorithms with regret bounds. In particular, for cardinality-based submodular functions, we give O(n 2)-time algorithms. © 2012 Springer-Verlag.

DOI： 10.1007/978-3-642-34106-9_22

Language：English   Publishing type：Research paper (other academic)  

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 (pm) 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.

DOI： 10.1007/978-3-642-24412-4_26

Event date： 2021.5

Language：Japanese  

Country：Japan  

Event date： 2021.5

Language：Japanese  

Country：Japan  

Event date： 2020.3

Language：English  

Venue：Snowmass Village   Country：United States  

We propose an online visual-object tracking method that is robust even in an adversarial environment, where various disturbances may occur on the target appearance, etc. The proposed method is based on a delayed-Hedge algorithm for aggregating multiple arbitrary online trackers with adaptive weights. The robustness in the tracking performance is guaranteed theoretically in term of "regret" by the property of the delayed-Hedge algorithm. Roughly speaking, the proposed method can achieve a similar tracking performance as the best one among all the trackers to be aggregated in an adversarial environment. The experimental study on various tracking tasks shows that the proposed method could achieve state-of-the-art performance by aggregating various online trackers.

Event date： 2019.11

Language：English  

Venue：Auckland   Country：New Zealand  

In this paper, we propose an optimal rejection method for rejecting ambiguous samples by a rejection function. This rejection function is trained together with a classification function under the framework of Learning-with-Rejection (LwR). The highlights of LwR are: (1) the rejection strategy is not heuristic but has a strong background from a machine learning theory, and (2) the rejection function can be trained on an arbitrary feature space which is different from the feature space for classification. The latter suggests we can choose a feature space which is more suitable for rejection. Although the past research on LwR focused only its theoretical aspect, we propose to utilize LwR for practical pattern classification tasks. Moreover, we propose to use features from different CNN layers for classification and rejection. Our extensive experiments of notMNIST classification and character/non-character classification demonstrate that the proposed method achieves better performance than traditional rejection strategies.

Event date： 2019.9

Language：English  

Venue：Sydney   Country：Australia  

In this paper, we analyze logo designs by using machine learning, as a promising trial of graphic design analysis. Specifically, we will focus on favicon images, which are tiny logos used as company icons on web browsers, and analyze them to understand their trends in individual industry classes. For example, if we can catch the subtle trends in favicons of financial companies, they will suggest to us how professional designers express the atmosphere of financial companies graphically. For the purpose, we will use top-rank learning, which is one of the recent machine learning methods for ranking and very suitable for revealing the subtle trends in graphic designs.

Event date： 2019.9

Language：English  

Venue：Sydney   Country：Australia  

Signature verification systems suffer from imbalanced learning, which imposes strict requirements on classifiers. The standard classification approaches, such as SVM, often degrade the performance for imbalanced data or require additional parameters for data balancing. In this study, as a new approach for signature verification, we use RankSVM as the writer-dependent classifiers, which theoretically guarantees the generalization performance for imbalanced data. To investigate the ability of RankSVM for solving imbalanced learning problems in signature verification tasks, the extensive experiments are conducted on bitmaps of GPDS-150, GPDS-300, GPDS-600, and GPDS-1000 datasets and deep features of GPDS-960 dataset. The experimental results demonstrate that the RankSVM-based approach obtains a nearly equivalent performance with the state-of-the-art method on deep features of the GPDS-960 dataset, and achieves significantly better performance than standard-SVM-based approach on bitmaps of GPDS-150, GPDS-300, GPDS-600, and GPDS-1000 datasets.

Event date： 2019.6

Language：English  

Venue：大分   Country：Japan  

Event date： 2019.6

Language：Japanese  

Venue：大分   Country：Japan  

Event date： 2017.7

Language：English  

Venue：Timisoara   Country：Other  

To discover teaching knowledge efficiently, we must extract the various teaching activities from educational data. In this paper, through the use of e-book logs and techniques of time-series analysis, we describe a method of practicing teaching analytics in face-to-face classes, one which enable us to extract the teaching activity efficiently and accurately.

Repository Public URL： http://hdl.handle.net/2324/4068613

Event date： 2017.7

Language：English  

Venue：Timisoara   Country：Other  

Students' reflective writings are useful not only for students themselves but also teachers. It is important for teachers to know which concepts were understood well by students and which concepts were not, to continuously improve their classes. However, it is difficult for teachers to thoroughly read the journals of more than one hundred students. In this paper, we propose a novel method to extract common topics and students' common impressions against them from students' journals. Weekly keywords are discovered from journals by scoring noun words with a measure based on TF-IDF term weighting scheme, and then we analyze co-occurrence relationships between extracted keywords and adjectives. We employs nonnegative matrix factorization, one of the topic modeling techniques, to discover the hidden impression topics from the co-occurrence relationships. As a case study, we applied our method on students' journals of the course 'Information Science' held in our university. Our experimental results show that conceptual keywords are successfully extracted, and four significant impression topics are identified. We conclude that our analysis method can be used to collectively understand the impressions of students from journal texts.

Event date： 2017.3

Language：English  

Venue：Vancouver   Country：Canada  

Many universities choose the C programming language (C) as the first one they teach their students, early on in their program. However, students often consider programming courses difficult, and these courses often have among the highest dropout rates of computer science courses offered. It is therefore critical to provide more effective instruction to help students understand the syntax of C and prevent them losing interest in programming. In addition, homework and paper-based exams are still the main assessment methods in the majority of classrooms. It is difficult for teachers to grasp students' learning situation due to the large amount of evaluation work. To facilitate teaching and learning of C, in this article we propose a system-LAPLE (Learning Analytics in Programming Language Education)-that provides a learning dashboard to capture the behavior of students in the classroom and identify the different difficulties faced by different students looking at different knowledge. With LAPLE, teachers may better grasp students' learning situation in real time and better improve educational materials using analysis results. For their part, novice undergraduate programmers may use LAPLE to locate syntax errors in C and get recommendations from educational materials on how to fix them.

Event date： 2012.10

Language：English  

Venue：Lyon   Country：France  

We consider an online prediction problem of combinatorial concepts where each combinatorial concept is represented as a vertex of a polyhedron described by a submodular function (base polyhedron). In general, there are exponentially many vertices in the base polyhedron. We propose polynomial time algorithms with regret bounds. In particular, for cardinality-based submodular functions, we give O(n ²)-time algorithms.

Repository Public URL： http://hdl.handle.net/2324/1932327

Event date： 2011.10

Language：English  

Venue：Espoo   Country：Finland  

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.

Repository Public URL： http://hdl.handle.net/2324/1546621

Event date： - 2021.5

Language：English  

Country：Japan  

Language：English   Presentation type：Symposium, workshop panel (public)  

Country：United States  

Language：Japanese  

Venue：東京   Country：Japan  

Language：Japanese  

Venue：沖縄   Country：Japan  

Language：Japanese  

Venue：東京   Country：Japan  

Language：English  

Online People-flow Prediction

Language：Others  

Learning theory and algorithms for shapelets and other local features

Language：Japanese  

Shapeletに基づいた文字認識

Language：Others  

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.

Language：English  

Time Series Classification Based on Random Shapelets

Language：Japanese  

Open Data in Education

Language：English  

Efficient AUC Maximization by Approximate Reduction of Ranking SVMs
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.

Language：English  

Online Prediction under Submodular Constraints
We consider an online combinatorial prediction problem where each combinatorial concept is represented as a vertex of a polyhedron described by a submodular function (base polyhedron). In general, there are exponentially many vertices in the base polyhedron. We propose polynomial time algorithms with regret bounds. In particular, for cardinality-based submodular functions, we give O(n^2)-time algorithms.

Language：English  

Approximate Reduction from AUC Maximization to 1-norm Soft Margin Optimization
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.

Language：Japanese  

Learning Shogi Evaluation Functions using Kernel Methods
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.

Language：Japanese  

Learning Evaluation Functions for Shogi Using SVM-based Bipartite Ranking Learning
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. To date, many complex features, such as relationships between multiple pieces, have been designed. In this paper, we propose an approach using polynomial kernels and Support Vector Machines (SVM), where only very simple features will be prepared explicitly. Polynomial kernels allow us to consider high dimensional, n-ary relations of monomial features. We further regard the problem of evaluation function learning as a bipartite ranking problem of the positions after legal moves, and propose a method which uses SVMs (ranking SVM). We show the effectiveness of our algorithm through computational experiments.

Type：Competition, symposium, etc. 

Type：Peer review 

Number of peer-reviewed articles in foreign language journals：1

Proceedings of International Conference Number of peer-reviewed papers：3

Proceedings of domestic conference Number of peer-reviewed papers：2