Academic

Energy-based probabilistic models are enjoying renewed popularity, given their success in applications to language and image processing. Modern methods for estimating and sampling these models overwhelmingly rely on annealing, which employs a sequence of intermediate distributions between a Gaussian and the data distribution. My research aims to identify the limitations of older methods, quantify how annealing improves their efficiency, and design principled strategies for selecting the intermediate distributions. In particular, we obtained theoretical guarantees for two popular sampling algorithms for diffusion models: annealed Langevin dynamics and time-reversed diffusions.