学習分類子システムのルール進化に対するConditional VAEに基づく誤判定検知・訂正

DOI

書誌事項

タイトル別名
  • Misclassification Detection and Correction based on Conditional VAE for Rule Evolution in Learning Classifier System

説明

<p>This paper proposes the Misclassification Detection and Correction method based on Conditional variational autoencoder (MDC/C) which detects and corrects the incorrect output of Learning Classifier System (LCS) through a comparison between the original data and the restored data by Conditional Variational Auto-Encoder (CVAE) with the output of LCS (as the condition to CVAE).The experimental results on the complex multi-class classification problem of the handwritten numerals have revealed the following implications: (1) although an integration of XCSR (i.e., the real-valued LCS) with CVAE (called CVAEXCSR) increases the correct rate in comparison with XCSR, it has the limit of improvement, i.e., the correct rate converges to 87.92%; (2) the correct rate of CVAEXCSR increases to 99.04% when removing the incorrect outputs by the detection mechanism of MDC/C and 95.03% when correcting them by the correction mechanism of MDC/C, respectively; and (3) the correct rate of CVAEXCSR with MDC/C is high from the first iterations and keeps it high even after the rule condensation which executes LCS without the crossover and mutation operations.</p>

収録刊行物

詳細情報 詳細情報について

  • CRID
    1390290769955326336
  • NII論文ID
    130008141304
  • DOI
    10.11394/tjpnsec.12.98
  • ISSN
    21857385
  • 本文言語コード
    ja
  • データソース種別
    • JaLC
    • CiNii Articles
  • 抄録ライセンスフラグ
    使用不可

問題の指摘

ページトップへ