A study on the mechanism of the minimum searching by the chaotic neural network
M. Ohta, Akio Ogihara, Shinobu Takamatsu, K. Fukunaga · 2002
This article analyzes dynamics of the chaotic neural network and its minimum searching principle. First, it is indicated that the dynamics of the chaotic neural network is described as a gradient descent method, and it is clarified that the behavior of the chaotic neural network can not only catch a local minimum of the energy but also escape from a local minimum without using any special technique. The performance of the chaotic behavior is then evaluated experimentally. In order to compare the chaotic behavior, a random minimum searching algorithm is provide. It is confirmed that chaos is more effective than random behavior from experimental results.