日本機械学会論文集 A編
Online ISSN : 1884-8338
Print ISSN : 0387-5008
ニューラルネットにおける重みの新しい決定法
森谷 之信天谷 賢治青木 繁
著者情報
ジャーナル フリー

1994 年 60 巻 570 号 p. 574-579

詳細
抄録
A new method to quickly determine weights in a layered neural network is introduced. By making use of the relation between the input of one unit in the (k+1) th layer and the output of all the units in the K th layer, all the weights can be obtained easily provided that the number of units in at least one layer is greater than or equal to the number of teacher sets plus 1. From two examples, it is shown that calculation time required by the new method is much shorter than that of back propagation method.
著者関連情報
© 社団法人日本機械学会
前の記事 次の記事
feedback
Top