Improving a Rule Induction System Using Genetic Algorithms

The field of automatic image recognition presents a variety of difficult classification problems involving the identification of important scene components in the presence of noise, changing lighting conditions, and shifting view points. This chapter describes part of a larger effort to apply machine learning techniques to such problems in an attempt to automatically generate and improve the classification rules required for various recognition tasks. The immediate problem attacked is that of texture recognition in the presence of noise and changing lighting conditions. In this context standard rule induction systems like AQ15 produce sets of classification rules which are not necessarily optimal with respect to: 1) the need to minimize the number of features actually used for classification, and 2) the need to achieve high recognition rates with noisy data. This chapter describes one of several multistrategy approaches being explored to improve the usefulness of machine learning techniques for such problems. The approach described here involves the use of genetic algorithms as a "front end" to traditional rule induction systems in order to identify and select the best subset of features to be used by the rule induction system. The proposed approach has been implemented and tested on difficult texture classification problems. The results are encouraging and indicate significant improvements can be obtained from a multistrategy approach in this domain.