Learning Restricted Boltzmann Machines

Friday, September 18, 2020 - 9:30am - 10:15am
Ankur Moitra (Massachusetts Institute of Technology)
Graphical models are a rich language for describing high-dimensional distributions in terms of their dependence structure. While there are algorithms with provable guarantees for learning undirected graphical models in a variety of settings, there has been much less progress in the important scenario when there are latent variables, which is the focus of our work.

In particular, we study Restricted Boltzmann Machines (RBMs) which are a popular model with wide-ranging applications. We gave a simple greedy algorithm for learning ferromagnetic RBMs. Our analysis is based on tools from mathematical physics that were developed to show the concavity of magnetization. Conversely we show that even for a constant number of latent variables with constant degree, without ferromagneticity the problem is as hard as sparse parity with noise. This hardness result is based on a sharp and surprising characterization of the representational power of bounded degree RBMs.

Based on joint work with Guy Bresler and Frederic Koehler