A pruning structure of self-organizing HCMAC neural network classifier

A self-organizing HCMAC neural network was proposed to solve high dimensional pattern classification problems well in our previous work. However, a large amount of redundant GCMAC nodes might be constructed due to the expansion approach of full binary tree topology. Therefore, this study presents a pruning structure of self-organizing HCMAC neural network to solve this problem. Experimental results show the proposed pruning structure not only can largely reduce memory requirement, but also keep fast training speed and has higher pattern classification accuracy rate than the original self-organizing HCMAC neural network does in the most testing benchmark data sets.