Learning Restricted Boltzman Machines

Channel:
United States Simons Institute for the Theory of Computing
Subscribers:
68,700
Published on ● Video Link: https://www.youtube.com/watch?v=eMEasYm_VWU


Duration: 0:00
751 views
0

Ankur Moitra (MIT)
https://simons.berkeley.edu/talks/tbd-394
Algorithmic Aspects of Causal Inference

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.



Other Videos By Simons Institute for the Theory of Computing


2022-03-23Causalsim: Trace-Driven Simulation For Network Protocols
2022-03-23Stability Of Causal Identification From The Perspective Of Condition Numbers
2022-03-23Confounding-Robust Policy Evaluation In Infinite-Horizon Reinforcement Learning
2022-03-23Identifying Mixtures Of Bayesian Network Distributions
2022-03-23Identifying Mixture Models
2022-03-22Efficient Distance Estimation And Causal Inference For Discrete Models
2022-03-22Parameter Estimation For Undirected Graphical Models With Hard Constraints
2022-03-22Causal Matrix Completion
2022-03-22Preconditioning In Sparse Linear Regression Using Graphical Structure
2022-03-22Learning Ill-Conditioned Gaussian Graphical Models
2022-03-22Learning Restricted Boltzman Machines
2022-03-22Neural Networks And Spurious Correlations
2022-03-21Tractable Probabilistic Circuits
2022-03-11The Impact of National Service on Beliefs, Mindsets, and Life Pathways...
2022-03-10Monitoring People and Their Vital Signs Using Radio Signals and Machine Learning
2022-03-08Causality and Autoencoders in Light of Drug Repurposing for COVID-19
2022-03-05Thinking Like a Journalist — Science Communicator in Residence Talk
2022-02-26Adversarial Machine Learning and Instrumental Variables for Flexible Causal Modeling
2022-02-26Online Adversarial Multicalibration And (Multi)Calibeating
2022-02-26Rebel: Combining Deep Reinforcement Learning and Search for Imperfect-Information Games
2022-02-26Generalized Energy-Based Models