Piecewise Deterministic Monte Carlo for Latent Variable Models - A Case Study

Explore a 49-minute research seminar where Dr. Joris Bierkens from Delft University of Technology discusses the application challenges and theoretical understanding of Piecewise Deterministic Monte Carlo (PDMC) methods in latent variable models. Delve into the attractive properties of PDMC, particularly its ability to use unbiased gradients of target distributions, and understand how this capability extends beyond data subsampling. Learn about recent developments in applying these properties to latent variable models, examining both the current challenges and valuable insights gained from studying this application. Part of the Stochastic Systems for Anomalous Diffusion series at the Isaac Newton Institute, this technical presentation contributes to the ongoing discourse in mathematical sciences and their practical applications.