專題討論 題目:Green Learning as A Renaissance of Classical Pattern Recognition 時間:11/09 14:00-15:30 地點:工程五館 A203室 講者:C.-C. Jay Kuo (University of Southern California)

Green Learning as A Renaissance of Classical Pattern Recognition

C.-C. Jay Kuo (University of Southern California)

November 9, 2026 (Mon.) 14:00~15:30 Room A203

Abstract: Classical pattern recognition (PR) comprises two stages: 1) feature extraction and 2) classification/regression. Features are developed ad hoc by experts with domain knowledge. Generally, features are more important than classifiers in performance. The resurgence of neural networks (NNs) in the 2010s has completely changed the AI/ML landscape. The end-to-end optimization of the parametric NN model blurs the boundary of the two PR stages. Over the past decade, I have investigated a new learning paradigm, called green learning (GL), that is competitive with deep learning in performance while significantly reducing model and computational complexities. In this talk, I will explain the relationship between GL and PR. In particular, I view GL as a renaissance of classical PR. Humans extract features, and output labels do not fine-tune them in PR. The data-fitting performance of PR is worse than that of neural-network-based deep learning because it lacks label feedback. GL enhances the discriminability of PR features using label-guided feature selection and new feature generation. I will use several examples to illustrate this point.

Bio:

Dr. C.-C. Jay Kuo received a Ph.D. from the Massachusetts Institute of Technology in 1987. He is now with the University of Southern California (USC) as the Ming Hsieh Chair Professor, a Distinguished Professor of Electrical and Computer Engineering and Computer Science, and the Director of the Media Communications Laboratory. He is a Fellow of the AAAS, ACM, IEEE, NAI, and SPIE, and an Academician of the Academia Sinica. He is a pioneer in green AI and its applications.