Weiming Feng
Tuesday, 19 March 2024, 16:00 - 17:00
ITS Seminar Room, Clausiusstrasse 47
Sampling Algorithms for High-Dimensional Distributions
High-dimensional distributions have been extensively studied in different research areas. Examples include spin systems in physics and Markov random fields in machine learning. Sampling is a central computational task for high-dimensional distributions, which requires the algorithm to generate random samples from the input distribution in polynomial time.
One of the most successful sampling algorithms is the Markov chain Monte Carlo (MCMC). I will introduce some results on theoretical analysis of MCMC algorithms. Recently, some new alternative sampling techniques were proposed. For example, resampling methods and sampling algorithms based on projection. The new techniques have some unique features compared to MCMC methods. I will also give some applications of sampling algorithms.