2 papers
stat.ML2025
Error Bounds and Optimal Schedules for Masked Diffusions with Factorized Approximations
Hugo Lavenant, Giacomo Zanella
Recently proposed generative models for discrete data, such as Masked Diffusion Models (MDMs), exploit conditional independence approximations to reduce the computational cost of p…
math.PR2024
Entropy contraction of the Gibbs sampler under log-concavity
Filippo Ascolani, Hugo Lavenant, Giacomo Zanella
The Gibbs sampler (a.k.a. Glauber dynamics and heat-bath algorithm) is a popular Markov Chain Monte Carlo algorithm which iteratively samples from the conditional distributions of…