Colloquia - Xuhui Fan, Uncertainties, Complexities and Scalable Inference: A Few Peeks at Bayesian Machine Learning, Virtual, 4:25 - 5:25 pm

Wednesday, December 8, 2021 - 4:25pm to 5:25pm
Event Type: 

Uncertainties, Complexities and Scalable Inference: A Few Peeks at Bayesian Machine Learning

Abstract:

By studying the joint probability of feature data (and label data), Bayesian Machine learning has become one of the most successful methods until now.
However, deficient modelling and inference issues have prevented the practical usages for notable models. In this talk, I will introduce three of my research
studies in the aspects of uncertainties, complexities and scalable inference: (1), nonparametric Bayesian space partition methods, which explore
various ways to generate partitions in a pre-defined space however are inevitable to generate noisy cuts; (2), multi-stochastic layered deep generative models, which
provide interpretable latent units in deep generative models however actually gain limited benefits from its deep structure, and (3), Gaussian Processes for time-dependent data, which uses flexible continuous functions to approximate the complex data however the existing approach is highly restricted. In addition to addressing these issues, my research made systematic developments for broader applications of these models. The talk will be concluded with brief discussions on the current challenges and future directions.

Bio:

Xuhui Fan received his bachelor's degree in mathematical statistics from the University of Science and Technology of China, Hefei, China, in 2010, and his Ph.D. degree in computer science from the University of Technology Sydney, Australia, in 2015. He then worked as a project engineer at Data61 (previously NICTA), CSIRO from 2015 to 2017. He is currently a Postdoc Fellow in the School of Mathematics and Statistics at the University of New South Wales, Sydney. He focuses on the topics of deep probabilistic models, Gaussian Processes, nonparametric Bayesian space partitioning methods, social network analysis, and publishes related research work in NeurIPS, ICML, AISTATS, etc.. He serves as senior PC members in IJCAI and AAAI and PC members in ICML, NeurIPS, AISTATS, ICLR.