Today, the neural network is used in many fields. In this study, chaotic neural network is used. It is closer to the work of a human brain. The learning method for this network is Incremental Learning method. Neurons learn according to its internal state on this learning method. The feature of this learning method is that there are different values of weight between two neurons. It is called asymmetry. So, symmetric weights from asymmetric weights are made with processing and the difference for these neural networks is searched.
The pattern of capital letter on alphabet is used for learning. Neural networks are made to learn this pattern, and symmetric weights are made with processing. Then networks are made to recall learned patterns. Consequently, some networks made symmetric weights by processing were not inferior to the network that have asymmetric weights. In the early learning, many symmetric weights network were better than asymmetric weights network.
Next, the kinds of learning pattern are made twice: capital letters and small letters. Learning ability declined on these networks. And, the process that moves weight greatly was tend to make the efficiency of learning lower. This result was brought through processing to make weights symmetry, not through symmetric weights. One network has ability equal to asymmetric weights of network.
Learning ability of Incremental Learning method is that it is gained by learning little by little,not by asymmetric weights.