Face Recognition by PDP and Radial Basis Function Networks: Comparisons and Insights

Shih-Cheng Yen, Paul Sajda, and Leif H. Finkel

Department of Bioengineering and
Institute of Neurological Sciences
University of Pennsylvania
Philadelphia, PA 19104, U. S. A.

syen@neuroengineering.upenn.edu
sajda@neuroengineering.upenn.edu
leif@neuroengineering.upenn.edu

Abstract

Despite a number of proposed neuropsychological and computational models, there is no accepted explanation for human face recognition abilities. We investigated the representations developed in both PDP and Radial Basis function networks presented with a large database of faces. Both networks achieved performances above 90% in gender classification tasks. Network representations were analyzed using a number of techniques including examination of connection weights, network inversion, ablation and modification of the image, and Wiener kernal - reverse correlation techniques. Comparison of the networks reveals a template-based strategy that combines statistical decision making with proto-type exemplars.

Full Text:
 PostScript
 Acrobat