ニューラルネットワークを用いた研削条件最適化に関する研究　追加学習方法について

DOI Web Site 被引用文献12件

書誌事項

タイトル別名
  • Study on Optimization of Grinding Conditions using Neural Networks. A Method of Additional Learning.
  • ニューラル ネットワーク オ モチイタ ケンサク ジョウケン サイテキカ ニ
  • A Method of Additional Learning
  • 追加学習方法について

この論文をさがす

抄録

An autonomous mechanical system with learning function, possessing capabilities for self-modification on the basis of empirically acquired knowledge and operating ability, was devised and implemented using a neural networks for the preparatory process in grinding operations. Applying above system, newly generated datum is acquired and successive additional learning is necessary. Formerly, in order to cope with this need, learning was repeated using all data, including the additional data. However restricted computational load and memory capacity required for learning increases vastly. Accordingly, this paper deals with a method whereby only recently acquired data and additional data, are learned, while the knowledge acquired through previous learning is preserved. Procedures were devised as follows. If the memory capacity is exceeded, then the datum with the greatest degree of similarity to the additional data is excluded from the memory and replaced, and clustering is employed for data compression in the memory area. In simulation experiments using random numbers, the present methods were compared with method such that the oldest datum was excluded from the memory area and additional learning is performed, and the results demonstrated that the present method decreases recall erros apprpximately one-half.

収録刊行物

  • 精密工学会誌

    精密工学会誌 58 (10), 1707-1712, 1992

    公益社団法人 精密工学会

被引用文献 (12)*注記

もっと見る

キーワード

詳細情報 詳細情報について

問題の指摘

ページトップへ