循環更新形式之神經網路

Abstract

在這篇論文之中，我們探討具短暫混沌性質的一種神經網路。我們將其中之對稱連接推廣到循環對稱連接。我們將討論此神經網路關於Lyapunov函數及其固定點的終極穩定性。在討論中，我們同時考慮了此神經網路之同步更新及循環更新的形式。比較這兩種更新形式的神經網路動態是很值得探討的課題。

This investigation extends the existence of Lyapunov function and asymptotic stability of fixed points for transiently chaotic neural networks from symmetric connectivity to cycle-symmetric connectivity. Both synchronous updating mode and cyclic updating mode for the networks are considered. It is of interest to compare the respective dynamics of the networks corresponding to these two different modes.