Professor Kar-Ann, Toh

School of Electrical and Electronic Engineering Yonsei University, Seoul, Korea Website: http://mi.yonsei.ac.kr/professor.html Email:

Biography

Kar-Ann Toh is a full professor in the School of Electrical and Electronic Engineering at Yonsei University, South Korea. He received the PhD degree from Nanyang Technological University (NTU), Singapore. He worked for two years in the aerospace industry prior to his post-doctoral appointments at research centres in NTU from 1998 to 2002. He was affiliated with Institute for Infocomm Research in Singapore from 2002 to 2005 prior to his current appointment in Korea. His research interests include biometrics, pattern classification, optimization and neural networks. He is a co-inventor of a US patent and has made several PCT filings related to biometric applications. Besides being an active member in publications (PAMI, Machine Learning, Neural Computation etc), Dr. Toh has served as a member of technical program committee for international conferences related to biometrics and artificial intelligence. He has also served as a reviewer for international journals including several IEEE Transactions. He is a senior member of the IEEE.

Title

Training a Neural Classifier Without Iteration

Abstract

In this talk, we shall first go through a brief account on important developments in neural network learning and subsequently zoom in to the use of neural networks for pattern classification. In view of the discrepancy between the frequently adopted least-squares error measure and the actual classification error count needed, we are motivated to seek a classification error formulation for direct cost minimization. Although the minimum classification error (MCE) problem has been a focused topic in the past few decades, a global and deterministic solution remained unattempted. By approximating the nonlinear counting step function using a quadratic link, we show that the classification error rate is deterministically solvable. Based on a single-layer feedforward network (SLFN), the proposal is supported with extensive numerical evidences.

Reference:
[1] K.-A. Toh, "Deterministic Neural Classification", Neural Computation, vol. 20, no. 6, pp.1565-1595, June 2008.