Feature Selection based on Genetic Algorithm compared to Mutual Information: A Case Study for Face Recognition
Aouatif Amine, Mohammed Rziza, Driss Aboutajdine · International journal of tomography and simulation · 2010
The goal of this paper is to study if there is a dependency between a selected feature vector at each generation of the genetic algorithm and the resulting fitness. In order to see the relation between these parameters, we first use DCT to transform each image as a feature vector (i.e., FFS). A GA is then used to select a subset of features from the low-dimensional representation by removing certain DCT coefficients that do not seem to encode important information about recognition task. When using SVM, two problems are confronted: how to choose the optimal input feature subset for SVM, and how to set the best kernel parameters. Therefore, obtaining the optimal feature subset and SVM parameters must occur simultaneously. We present a genetic algorithm approach for feature selection and parameters optimization to solve this kind of problem. Our proposed algorithm was compared with CMIM, IGFS and mRMR criteria. The experiment results indicate the robustness of our work.