In recent years, deep generative models have been largely dismissed for fully-supervised classification settings as they are often substantially outperformed by discriminative deep classifiers (e.g., softmax classifiers). In contrast to this common belief, we show that it is possible to formulate a simple generative classifier that is significantly more robust to (a) detecting out-of-distribution samples (i.e., novelty) and (b) training noisy labeled data, without much sacrifice of the original discriminative performance with respect to in-distribution or/and clean labeled data. This is a joint work with Kimin Lee (KAIST), Kibok Lee (U. of Michigan) and Honglak Lee (U. of Michigan).
Jinwoo Shin obtained B.S. degrees (in Math and CS) from Seoul National University in 2001, and the Ph.D. degree (in Math) from Massachusetts Institute of Technology in 2010 with George M. Sprowls Award (for best MIT CS Ph.D. theses). He was a postdoctoral researcher at Algorithms & Randomness Center, Georgia Institute of Technology in 2010-2012 and Business Analytics and Mathematical Sciences Department, IBM T. J. Watson Research in 2012-2013. He joined KAIST in Fall 2013 and is currently an associate professor at the School of Electrical Engineering at KAIST. Dr. Shin has worked on developing advanced algorithms for machine learning and stochastic networks, under a mixture of flavors including applied probability and theoretical computer science. He received the Rising Star Award in 2015 from the Association for Computing Machinery (ACM) Special Interest Group for the computer systems performance evaluation community (SIGMETRICS). He also received the Kenneth C. Sevcik Award at ACM SIGMETRICS/Performance 2009, Best Publication Award from INFORMS Applied Probability Society 2013, Best Paper Award at ACM MOBIHOC 2013 and Bloomberg Scientific Research Award 2015. He has published papers at top conferences of various disciplines: NIPS/NeurIPS, ICML, ICLR, UAI, AISTATS (machine/deep learning), CVPR (computer vision), SIGMETRICS, INFOCOM, MOBIHOC (networks), FOCS (theoretical computer science), ISIT (information theory), OSDI (operating system), S&P (security) and CDC (control).