/RBM

Representing the 2D Ising model with a Restricted Boltzmann Machine

Primary LanguageJupyter NotebookMIT LicenseMIT

RBM

Representing the 2D Ising model with a Restricted Boltzmann Machine

The Restricted Boltzmann Machine is trained to represent the probability distribution of the 2D Ising Model. It is trained on data generated by the Metropolis-Hastings algorithm. The Kullback-Leibler divergence between the probability distributions is minimized using stochastic gradient descent and Gibb's sampling.

This code was used in a project completed with Dr. Ehsan Khatami