PARALLEL SOLUTION OF FEATURES SUBSET SELECTION PROCESS FOR HAND-PRINTED CHARACTER RECOGNITION

Abstract

The existence of a huge amount of features for pattern recognition problems brings to the overloading of the training and exploitation steps of the recognition; also, highly correlated features affect the accuracy of the designed systems negatively. One of the most used ways for tackling this problem is the application of genetic algorithms for the solution of the binary optimization problems that appeared during the features subset selection process. In this paper was used parallel genetic algorithms for the selection of the most informative features in Azerbaijani hand-printed character recognition system by using opportunities of the distributed cluster computing. In this way after the given number of generations most appropriate features with the high recognition rate were selected from the features database.