Improving BPSO-based feature selection applied to offline WI handwritten signature verification through overfitting control

This paper investigates the presence of overfitting when using Binary\nParticle Swarm Optimization (BPSO) to perform the feature selection in a\ncontext of Handwritten Signature Verification (HSV). SigNet is a state of the\nart Deep CNN model for feature representation in the HSV context and contains\n2048 dimensions. Some of these dimensions may include redundant information in\nthe dissimilarity representation space generated by the dichotomy\ntransformation (DT) used by the writer-independent (WI) approach. The analysis\nis carried out on the GPDS-960 dataset. Experiments demonstrate that the\nproposed method is able to control overfitting during the search for the most\ndiscriminant representation.\n