Noise Effects and Fault Tolerance in Backpropagation Learning

  • Morie T.
    Faculty of Engineering, Hiroshima University
  • Nakamura T.
    Faculty of Engineering, Hiroshima University
  • Nagata M.
    Faculty of Engineering, Hiroshima University
  • Iwata A.
    Faculty of Engineering, Hiroshima University

Bibliographic Information

Other Title
  • 誤差逆伝搬学習におけるノイズの効果とフォールトトレランス

Search this article

Description

This paper describes effects of noise injection into the backpropagation learning process in analog neural network circuits with relation to synaptic weight quantization, fault tolerance, and applicability of chaotic noise instead of random noise. Most simulation results suggest that weight quantization and noise injection affect the learning performance independently. Other simulation results demonstrate that logistic chaotic noise injection has the same effect on improving fault tolerance of backpropagation networks as random noise injection.

Journal

  • Technical report of IEICE. FTS

    Technical report of IEICE. FTS 98 (68), 61-67, 1998-05-22

    The Institute of Electronics, Information and Communication Engineers

Keywords

Details 詳細情報について

  • CRID
    1570854177376004992
  • NII Article ID
    110003194147
  • NII Book ID
    AN10012998
  • Text Lang
    ja
  • Data Source
    • CiNii Articles

Report a problem

Back to top