Dynamic ensemble selection for off-line signature verification

Dynamic ensemble selection for off-line signature verification

Batista, Luana and Granger, Eric and Sabourin, Robert

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 2011

Abstract : Although not in widespread use in Signature Verification (SV), the performance of SV systems may be improved by using ensemble of classifiers (EoC). Given a diversified pool of classifiers, the selection of a subset to form an EoC may be performed either statically or dynamically. In this paper, two new dynamic selection (DS) strategies are proposed, namely OP-UNION and OP-ELIMINATE, both based on the K-nearest-oracles. To compare ensemble selection strategies, a hybrid generative-discriminative system for off-line SV system is considered. Experiments performed by using real-world SV data, comprised of genuine samples, and random, simple and skilled forgeries, indicate that the proposed DS strategies achieve a significantly higher level of performance in off-line SV than other well-known DS and static selection (SS) strategies. Improvements are most notable in problems where a significant level of uncertainty emerges due a considerable amount of intra-class variability. © 2011 Springer-Verlag.