Learning and bifurcation diagram of recurrent neural networks

被引：0

作者：

Sato, S ^{[1
]}

Gohara, K ^{[1
]}

机构：

[1] Hokkaido Univ, Dept Appl Phys, Fac Engn, Sapporo, Hokkaido 060, Japan

来源：

PROGRESS IN CONNECTIONIST-BASED INFORMATION SYSTEMS, VOLS 1 AND 2 | 1998年

关键词：

D O I：

暂无

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

The-input-output relation of recurrent neural networks are analyzed by using a bifurcation diagram of dynamical systems. We show experimentally that network behavior is represented as an approximate transition between excited attractors on a bifurcation diagram.

引用

页码：363 / 366

页数：4