Pattern classification using fuzzy sets and neural nets: A case-based approach

The present article describes a case-based pattern classification method in connectionist framework using fuzzy set theory. Some labeled samples from each class are selected automatically as cases based on the concept of fuzzy similarity, and are represented as hidden nodes. The number of nodes is determined by the extent of fuzzy regions around cases. These are determined adaptively through growing and pruning under supervised training of the network. The effectiveness of the system, along with comparisons, has been demonstrated on various synthetic and real life pattern recognition problems for different extent of fuzzy regions.