Quantitative Workshop: Latent Variable Models and Inference Methods

This workshop aims to introduce students to latent variable models and the statistical methods used to infer hidden structure in complex datasets. Topics will include mixture models, factor analysis, time series models such as hidden Markov models (HMMs), and modern nonlinear deep generative models such as variational autoencoders (VAEs). We will discuss core inference techniques including the Expectation-Maximization (EM) algorithm and variational inference. If time permits, we will discuss related concepts from information theory and Bayesian inference. The workshop is intended for Ph.D. students with quantitative backgrounds (e.g., neuroscience, engineering, physics, computer science, statistics, or psychology) who are interested in probabilistic modeling and data analysis. The workshop will be followed by a trainee dinner, and there will also be time on the morning of December 5th for individual or group meetings with students.