Soft Computing Tools and Pattern Recognition
Sankar Kumar Pal · IETE Journal of Research · 1998
Relevance of different soft computing tools e.g., fuzzy sets, artificial neural networks and genetic algorithms to pattern recognition problems is explained. Their distinguishing characteristics and roles in the soft computing framework are stated. Genetic algorithmic approach, being relatively new to the pattern recognition community, is paid more attention. A classification methodology based on this approach is described in detail along with its different features and a comparison in performance with the related methods. The effect of incorporating variable string length and differentiation in chromosome is discussed. Relation with Bayes decision boundary and analogy with multilayer perceptron based classification are explained. Finally, the merits of integrating the different soft computing tools for designing an efficient decision making system are stated with some application specific examples on neuro-fuzzy and neurogenetic approaches. Scope for further research is outlined. An extensive bibliography is also provided.