- This event has passed.
PRiML Seminar: “Optimizing probability distributions for learning: sampling meets optimization”
February 22 at 3:00 PM - 4:00 PM
Optimization and sampling are both of central importance in large-scale machine learning problems, but they are typically viewed as very different problems. This talk presents recent results that exploit the interplay between them. Viewing Markov chain Monte Carlo sampling algorithms as performing an optimization over the space of probability distributions, we demonstrate analogs of Nesterov’s acceleration approach in the sampling domain, in the form of a discretization of an underdamped Langevin diffusion. In the other direction, we view stochastic gradient optimization methods, such as those that are common in deep learning, as sampling algorithms, and study the finite-time convergence of their iterates to an invariant distribution.
Professor of Computer Science and Statistics, University of California at Berkeley
Peter Bartlett is a professor in the Computer Science Division and Department of Statistics and Associate Director of the Simons Institute for the Theory of Computing at the University of California at Berkeley. His research interests include machine learning and statistical learning theory. He is the co-author, with Martin Anthony, of the book Neural Network Learning: Theoretical Foundations. He has served as an associate editor of the journals Bernoulli, Mathematics of Operations Research, the Journal of Artificial Intelligence Research, the Journal of Machine Learning Research, and the IEEE Transactions on Information Theory, and as program committee co-chair for COLT and NIPS. He was awarded the Malcolm McIntosh Prize for Physical Scientist of the Year in Australia in 2001, and was chosen as an Institute of Mathematical Statistics Medallion Lecturer in 2008, an IMS Fellow and Australian Laureate Fellow in 2011, and a Fellow of the ACM in 2018. He was elected to the Australian Academy of Science in 2015.