Piecewise Deterministic Markov Process

Piecewise Deterministic Markov Processes (PDMPs) are continuous-time stochastic processes used to model systems exhibiting both deterministic and random behavior, offering advantages over traditional Markov Chain Monte Carlo methods in various applications. Current research focuses on developing efficient PDMP-based algorithms for Bayesian inference in complex models like neural networks, and for addressing challenges in reinforcement learning, particularly in non-stationary and multi-agent environments. These advancements improve sampling efficiency, enabling scalable inference and more robust decision-making in diverse fields, including high-dimensional volume computation and online learning.