概要:

In Incremental Learning on the chaotic neuron network, the time attenuation constants which are the parameter of the chaotic neuron can be 0, when input patterns do not include noise.

In this study, how the chaotic neural netowork works was inspected when the time attenuation constants are 0. The maximum number of complete learning was observed when $\alpha$ and $\mathit {\Delta } w$ changed.

As a result, it was confirmed that learning result changed by $\alpha$ and $\mathit {\Delta } w$. Learning ability was the highest around $\alpha = 0.245$ and $\mathit{\Delta} w = 0.00010$ when chaotic neural network was given input patterns which do not include noise. The maximum number of complete learning at that time was 162.

Next, how this chaotic neural network works when another patterns were input. These patterns are the ones with which the maximun number of complete learning was 149 when the time attenuation constants were not 0.

The result showed that the maximum number of complete learning was 154.

As a result of experiments, the learning ability was higher when the time attenuation constants were 0.