Today, the neural network is used in many fields. In 1943, the first study of neural network was begun by McCulloch and Pitts. Now, study of Chaotic Neural Network which is introduced a chaos element become active.
By the way, there is Incremental Learning in chaotic neural networks learnings. Incremental learning is applied the Hebbian rule.
Using incremental learning, the weights either increase or decrease more than necessarily. Then, in this study, the weights are standardized. The standardization is that the weights are divided according to previously determined the Standard Maximum Value.
The study clarifies that the quantity of patterns which can be learned is restricted by the network, when wight standardizes. The network has currently few patterns learned when standard maximum value is small.
It was also shown that the quantity of patterns which can be learned was influenced by the refectory scaling parameter. Using the same standard maximum value, the network with smaller refectory scaling parameter is able to learn more patterns.
と標準最大値