主讲人：黄坚 Professor,University of Iowa
地 点：线上Zoom会议ID: 86860610732密码：721918
Learning a probability distribution based on a random sample and sampling from a given distribution are two of the fundamental problems in statistics and machine learning. There two problems have been studied intensively as two separate questions in the literature. For example, many methods have been developed for nonparametric density estimation; likewise, many types of MCMC algorithms have been introduced for sampling from (unnormalized) distributions in Bayesian statistics. In recent years, generative learning approaches such as GANs have been proven effective in learning distributions of high-dimensional complex data. In this talk, we demonstrate that deep generative learning can also be adapted for sampling from unnormalized distributions. We present a relative entropy gradient sampler (REGS) for sampling from unnormalized distributions. This is a generative learning approach implemented using Wasserstein gradient flows. The key to the success of such an approach is the power of deep neural networks in approximating high-dimensional functions. Extensive simulation experiments on challenging multimodal mixture distributions and Bayesian logistic regression on real datasets demonstrate that the REGS outperforms the state-of-the-art sampling methods included in the study.
Professor Huang’s research interests include high-dimensional statistics, machine learning, survival analysis, bioinformatics and statistical genetics. He has published over 100 peer-reviewed papers. Many of his publications appeared in top-ranked journals, including Annals of Statistics, Bioinformatics, Biometrics, Biometrika, Econometrika, Journal of the American Statistical Association, Journal of Machine Learning Research, PNAS, and The American Journal of Human Genetics. Professor Huang was bestowed a National Institutes of Health Research Scientist Development Award in 1998. He is a Fellow of the American Statistical Association and the Institute of Mathematical Statistics. He served as associate editor of Annals of Statistics, Statistica Sinica and Statistics and Its Interface. Professor Huang has been designated a Highly Cited Researcher from 2015 to 2019 by the Web of Science group, ranking among the top one percent of researchers from 2003 to 2019, for most cited papers in the field of Mathematics.