Parallel Session [T8-09B]

[T8-09B-03] Impact of Short Non-Convergent MCMC Samplings on training Energy-Based Generative Models

Beatriz Seoane^1,2, Elisabeth Agoritsas³, *Giovanni Catania², Aurélien Decelle^2,1 (1. Université Paris-Saclay, CNRS, INRIA Tau team, LISN, 91190, Gif-sur-Yvette, France., 2. Departamento de Fisica Teorica, Universidad Complutense de Madrid, 28040 Madrid, Spain., 3. Department of Quantum Matter Physics, University of Geneva, 1211 Geneva, Switzerland. )

Keywords:Generative methods, Inference problems, Energy-based models, Monte Carlo simulations, Unsupervised Machine Learning