研究者詳細 - 中田　雅也

<p>The XCSAM classifier system is an approach of evolutionary rule-based machine learning, which evolves rules advocating the highest-return actions at state, resulting in best classification. This paper starts with claiming a limitation that XCSAM still fails to evolutionary generate adequate rules advocating the highest-return actions. Then, under our hypothesis that this limitation is caused from the rule-deletion mechanism of XCSAM, we revisit its rule-deletion strategy in order to promote the evolution of adequate rules. Different from the existing deletion strategy which deletes two rules for each rule-evolution, our deletion strategy is modified to delete more than two rules as necessary. Experimental results on a benchmark classification task validate our modification powerfully works to evolve adequate rules, improving the performance of XCSAM. We further show our modification robustly enables XCSAM to perform well on real world classification task.</p>

An Analysis of Rule Deletion Scheme in XCS on Reinforcement Learning Problem

Masaya Nakata, Tomoki Hamagami

Journal of Advanced Computational Intelligent Information 21 ( 5 ) 876 - 884 2017年5月 [査読有り]

DOI CiNii Research

記述言語：英語掲載種別：研究論文（学術雑誌）出版者・発行元：富士技術出版共著

<p>The XCS classifier system is an evolutionary rule-based learning technique powered by a Q-learning like learning mechanism. It employs a global deletion scheme to delete rules from all rules covering all state-action pairs. However, the optimality of this scheme remains unclear owing to the lack of intensive analysis. We here introduce two deletion schemes: 1) local deletion, which can be applied to a subset of rules covering each state (a match set), and 2) stronger local deletion, which can be applied to a more specific subset covering each state-action pair (an action set). The aim of this paper is to reveal how the above three deletion schemes affect the performance of XCS. Our analysis shows that the local deletion schemes promote the elimination of inaccurate rules compared with the global deletion scheme. However, the stronger local deletion scheme occasionally deletes a good rule. We further show that the two local deletion schemes greatly improve the performance of XCS on a set of noisy maze problems. Although the localization strength of the proposed deletion schemes may require consideration, they can be adequate for XCS rather than the original global deletion scheme.</p>

許容誤差を自己適応可能な学習分類子システム

辰巳嵩豊, 小峯嵩裕, 中田雅也, 佐藤寛之, 髙玉圭樹

進化計算学会論文誌 6 ( 2 ) 90 - 103 2016年4月 [査読有り]

DOI CiNii Research

記述言語：日本語掲載種別：研究論文（学術雑誌）出版者・発行元：進化計算学会共著

The XCS classifier system is designed to evolve accurately generalized classifiers as an optimal solution to a problem. All classifiers are identified as either accurate or inaccurate on the basis of a pre-defined parameter called an accuracy criterion. Previous results suggested a standard setting of the accuracy criterion robustly performs on multiple simple problems so XCS evolves the optimal solution. However, there lacks a guideline of reasonable setting of accuracy criterion. This causes a problem that the accuracy criterion should be empirically customized for each complex problems especially noisy problems which is a main focus of this paper. This paper proposes a self-adaptation technique for the accuracy criterion which attempts to enable XCS to evolve the optimal solution on the noisy problems. In XCS-SAC(XCS with Self-Adaptive accuracy criterion), each classifier has its own accuracy criterion in order to find an adequate setting of accuracy criterion for each niche. Then, each classifier's accuracy criterion is updated with the variance of reward which its classifier has received. We test XCS-SAC on a benchmark classification problem (i.e., the multiplexer problem) with noise (the Gaussian noise and alternative noise). Experimental results show XCS-SAC successfully solves the noisy multiplexer problems as well as XCS but evolves a more compact solution including an optimal solution than XCS.

XCS-SL: A Rule-based Genetic Learning System for Sequence Labelling

Nakata, Masaya and Kovacs, Tim and Takadama, Keiki

Evolutionary Intelligence 8 ( 2 ) 133 - 148 2015年9月 [査読有り]

DOI

記述言語：英語掲載種別：研究論文（学術雑誌）出版者・発行元：Springer 共著

Rule Reduction by Selection Strategy in XCS with Adaptive Action Map

Nakata, Masaya and Pier Luca Lanzi and Takadama, Keiki

Evolutionary Intelligence 9 ( 2 ) 71 - 97 2015年9月 [査読有り]

DOI

記述言語：英語掲載種別：研究論文（学術雑誌）出版者・発行元：Springer 共著

Evolutionary Algorithm for Uncertain Evaluation Function, New Mathematics and Natural Computation,

Tajima, Yusuke and Nakata, Masaya and Matsushima, Hiroyasu and Ichikawa, Yoshihiro and Sato, Hiroyu … 全著者表示

New Mathematics and Natural Computation 11 ( 2 ) 201 - 215 2015年2月 [査読有り]

DOI

記述言語：英語掲載種別：研究論文（学術雑誌）出版者・発行元：World Scientific 共著

Compact Genetic Algorithmを導入した学習分類子システムによる分類子数の削減

中田雅也, Pier Luca Lanzi, 田島友祐, 高玉圭樹

情報処理学会論文誌：数理モデル化と応用 7 ( 2 ) 1 - 16 2014年11月 [査読有り]

記述言語：日本語掲載種別：研究論文（学術雑誌）出版者・発行元：The Information Processing Society of Japan 共著

予測報酬に基づく個別化による学習分類子システムの学習性能の向上

中田雅也, 原田智広, 佐藤圭二, 松島博康, 高玉圭樹

計測自動制御学会論文集 48 ( 11 ) 713 - 722 2012年11月 [査読有り]

DOI CiNii Research

記述言語：日本語掲載種別：研究論文（学術雑誌）出版者・発行元：Society of Instrument and Control Engineers 共著

This paper focuses on Identification-based XCS (IXCS) which introduces the identification mechanism into XCS (Accuracy-based Leraning Classifier System) and extends it to Predicted reward-based IXCS(PIXCS) to promote a generalization of classifiers(i.e., rules) in the binary and multi-classification problems with reducing the number of classifiers. Through the intensive simulation of 20-Multiplexer problem and 3×3 Concatenated multiplexer problem, this paper has revealed the following implications which cannot be achieved by the conventional LCS(i.e., XCSTS) and IXCS: (1) PIXCS can derive better performance than XCSTS and IXCS in the binary-classification problem, (2) PIXCS can generalize not only the classifiers faster than IXCS but also the classifiers which are robust in the noisy multi-classification problem with reducing the number of classifiers.

個別化による学習分類子システムの一般化促進

中田雅也, 原田智広, 佐藤圭二, 松島裕康, 高玉圭樹

計測自動制御学会論文集 47 ( 11 ) 581 - 590 2011年11月

DOI CiNii Research

記述言語：日本語掲載種別：研究論文（学術雑誌）出版者・発行元：計測自動制御学会単著

This paper proposes a novel Learning Classifier System (LCS) called Identification-based LCS (IXCS) to promote a generalization of classifiers (i.e., rules) by selecting effective ones and deleting ineffective ones. Through the intensive simulation of the 20-Multiplexer problem, this paper has revealed the following implications which cannot be achieved by the conventional LCS, XCSTS: (1) IXCS can not only generalize the classifiers earlier but also generate the classifiers which are robust to the noisy environment; and (2) IXCS can derive a higher performance with a lower number of micro-classifiers.

このページの先頭へ▲

<中田 雅也>

論文 - 中田 雅也

<中田　雅也>

論文 - 中田　雅也