Ieee Transactions on Signal Processing

|Two training algorithms for self evolving neural networks are discussed for rule based data analysis. E cient classi cation is achieved with less number of automatically added clusters and application data is analysed by interpreting the trained neural network as a fuzzy rule based system. Learning Vector Quantisation algorithm has been modied acquiring the self evolvement character in the prototype neuron layer based on sub Bayesian decision making. The number of required prototypes representing fuzzy rules is automatically determined by the application data set. This method, compared with others shows better classi cation results for data sets with high noise or overlapping classi cation boundaries. The classifying Radial Basis Function Networks are generalised into Multiple Shape Basis Function Networks. The learning algorithm discussed is capable of adding new neurons representing self evolving clusters of di erent shapes and sizes dynamically. This shows a clear reduction in number of neurons or the number of fuzzy rules generated and the classi cation accuracy is increased signi cantly. This improvement is highly relevant in developing neural networks functionally equivalent to fuzzy classi ers since the transparency is strongly related to the compactness of the system. Keywords|Multiple Shape Basis Functions, Fuzzy rules, Clustering, Learning Vector Quantisation, Radial Basis Function Networks