Full Title:
A Short Introduction to Variational Bayesian Inference

Abstract:
This is a tutorial on the variational inference approach. One of the core problems in modern statistical learning is to approximate difficult-to-compute probability densities. Particularly in Bayesian statistics, posterior distributions of the unknown parameters and predictions need to be calculated, and they are difficult to obtain for complex models. Variational Inference is widely used to approximate the posterior densities for Bayesian models, as an alternative strategy to Markov chain Monte Carlo (MCMC) sampling. In this tutorial, we introduce the mean-field variational inference, which is a relatively simple version of the many VI methods, and show how it can be used to different statistical learning methods. We will also learn some other nonparametric and more advanced variational inference methods.

Short Bio:
Dr. Lulu Kang is an Associate Professor of the Department of Applied Math at Illinois Institute of Technology. She holds an M.S. in Operations Research and a Ph.D. in Industrial Engineering from Georgia Institute of Technology. Dr. Kang’s research focus is data science. Specifically, her research areas include statistical learning, uncertainty quantification, statistical design and analysis of experiments, Bayesian computational statistics, optimization and their application in complex systems in manufacturing, energy, and other engineering fields. Dr. Kang serves as the associate editor for journals SIAM/ASA Journal on Uncertainty Quantification and Technometrics. Dr. Kang created the Master of Data Science program at Illinois Tech and has been running the program since 2013. She has also created new curriculums and courses related data science.