Framework for evaluating quantum-assisted training of Restricted Boltzmann Machines

Restricted Boltzmann Machines (RBM) are neural networks that can be used to learn underlying data distributions. RBMs fall within the class of Energy Based Models, where the dependence between their variables can be represented by an energy function E(v, h). In order to train this model, different methods have been proposed, with the common goal of approximating the variable distributions, such as Contrastive Divergence and Persistent Contrastive Divergence. In this work, we are investigating different ways that Quantum Computing (QC) or Quantum-Inspired technologies can aid the training process. To accomplish that, a framework that implements different quantum-assisted training techniques is being developed, leveraging an interface with contrasting QC platforms.