Limited scope learning for self-organizing map and its applications

Self-Organizing Map (SOM) is a kind of neural network that teams without supervision. In this paper, we propos "Limited Scope Learning" on SOM. This technique is able to get the feature map that is shown by distributed expression, i.e., more than one winner is selected out of the whole map at the learning time. In the case that the troubled nodes exist on the map, the degree of node fault tolerance will be improved by using this method rather than the conventional technique.