Applying Fuzzy Clustering Based Neural Network to Life and Death in Go
Hongxu Ma · Jisuanji fangzhen · 2007
The life and death problems in Go consume too much resource and the number of nodes in the game tree increases exponential by the branching factor and depth of the game tree,which makes traditional search methods invalidated.In this paper a fuzzy clustering based neural network which uses pattern recognition and fuzzy properties detection is presented.A neural network based pattern clustering analyzer is constructed for the life and death problems which involve extern seki,hash permutation,circulation elusion and prove tree.With this approach,the branching factor of the game tree is greatly reduced,which saves enormous computing time and memory space.Experiment shows that the result is satisfactory and proves that applying self-study ability to programs is an efficient approach to improve the performance of computer Go.