Summary
International Symposium on Nonlinear Theory and Its Applications
2016
Session Number:B3L-C
Session:
Number:B3L-C-1
Stability Analysis of Periodic Orbits in Dynamic Binary Neural Networks
Shunsuke Aoki, Ryuji Sato, Kazuma Makita, Toshimichi Saito,
pp.-
Publication Date:2016/11/27
Online ISSN:2188-5079
DOI:10.34385/proc.48.B3L-C-1
PDF download (406.6KB)
Summary:
This paper studies various periodic orbits and their stability in dynamic binary neural networks. The networks are characterized by the signum activation function and ternary connection parameters. Depending on parameters, the network can generate a binary periodic orbit. The dynamics is simplified into a digital return map defined on a set of points. In order to analyze the stability of the periodic orbits, we present a feature plane of simple feature quantities. Calculating the feature quantities for a simple class of networks, we investigate variety of the periodic orbits, stability of them, and super-stability of them.