Speaker
Joseph Hadley
(University of Liverpool)
Description
The use of generative models to learn and sample complex distributions is increasingly common in computational physics. Many generative approaches are being used with a view to improving algorithms for complex lattice systems like QCD. One such generative model is the Variational Autoencoder, which can simplify a complex distribution by identifying the distribution with a Gaussian in a latent space. In this work we use a Variational Autoencoder to learn efficient Monte-Carlo updates to the Ising model. We use this model as an example to discuss ergodicity and detailed balance conditions within the Metropolis-Hastings algorithm.
Primary author
Joseph Hadley
(University of Liverpool)
