Langevin Algorithms in Machine Learning

*Presentation will be recorded.

Abstract: 

Langevin algorithms are core Markov Chain Monte Carlo methods for solving machine learning problems. These methods arise in several contexts in machine learning and data science including Bayesian learning problems with high-dimensional models and stochastic non-convex optimization problems including the challenging problems arising in deep learning. In this talk, we first discuss Langevin algorithms for non-convex optimization, and show that acceleration is possible for momentum-based Langevin algorithms in the context of non-convex optimization. Next, we illustrate the applications of Langevin algorithms in sampling and Bayesian learning via decentralized Langevin algorithms and penalized Langevin algorithms. Finally, we discuss heavy-tailed Langevin dynamics to approximate stochastic gradient algorithms and provide insights in the context of deep learning, as well as related algorithms and phenomena.

Bio: 

Lingjiong Zhu got his BA from University of Cambridge in 2008 and PhD from New York University in 2013. He worked at Morgan Stanley and University of Minnesota before joining the faculty at Florida State University in 2015, where he is currently an Associate Professor. His research interests include applied probability, data science, and operations research. His works have been published in many leading outlets including Annals of Applied Probability, Operations Research, Production and Operations Management, INFORMS Journal on Computing, ICML, NeurIPS and Journal of Machine Learning Research. His research has been supported by three NSF grants and a Simons Collaboration Grant. He was a recipient of Kurt O. Friedrichs Prize for an outstanding dissertation from Courant Institute, New York University in 2013 and Developing Scholar Award from Florida State University in 2022.