Analog filter circuits feature selection using MRMR and SVM

This paper addresses frequency response feature extraction of analog filter circuits. An approach for feature selection using criteria of maximum relevance minimum redundancy (MRMR) and support vector machine (SVM) is proposed. Key idea of the method is to obtain candidate feature subsets with descending order using criteria of MRMR first, and then to select the optimal feature subset through cross validation results of each feature subset using SVM. Experimental results testify that the proposed approach has the advantages of less time-consuming and the selected features have reasonable distribution.