The purpose of this study is to simplify the chaotic neurons in the incremental learning and to shorten the time the chaotic neural network takes to learn. The time summation terms are omitted to simplify the chaotic neurons.
The time summation term is a term for obtaining the time summation in the input and refractory.
First, the influence was examined after omitting the time summation term. The period the network is learning was shortened after the time summation terms were omitted.
Next, the relationship between The number of learned patterns and the number of times of pattern input was examined. The number of times of pattern input was 100 at the former examinations. It could be reduced to about 20 if there is the time summation term. It could be reduced to about 3 if there is no the time summation term.
In addition, influence of noise was examined. If the time summation term omitted, the number of learned patterns decreased due to the influence of noise.
Finally, the parameters were explored to omit the number of times of pattern input. The parameter by which influence of noise is small and by which the number of times of pattern input could be reduced was found. But, the number of learned patterns reduced.
When there is no noise on the input, it is possible to shorten the learning time by omitting the time summation term.