Feature subset selection using an optimized hill climbing algorithm for handwritten character recognition

Feature subset selection using an optimized hill climbing algorithm for handwritten character recognition

Nunes, Carlos M. and De Britto, Alceu S. and Kaestner, Celso A.A. and Sabourin, Robert

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 2004

Abstract : This paper presents an optimized Hill Climbing algorithm to select a subset of features for handwritten character recognition. The search is conducted taking into account a random mutation strategy and the initial relevance of each feature in the recognition process. The experiments have shown a reduction in the original number of features used in an MLP-based character recognizer from 132 to 77 features (reduction of 42%) without a significant loss in terms of recognition rates, which are 99% for 60,089 samples of digits, and 93% for 11,941 uppercase characters, both handwritten samples from the MIST SD19 database. The proposed method has shown to be an interesting strategy to implement a wrapper approach without the need of complex and expensive hardware architectures. ©Springer-Verlag 2004.