Learning and Evaluating Boltzmann Machines

By Ruslan Salakhutdinov. 2008

74 citations recorded by the source corpus.

Subfields: Probabilistic & Bayesian Methods

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Probabilistic & Bayesian Methods

Citations in the corpus

Barrier Frank-Wolfe for Marginal Inference (2015, background): “RBMs: As in Salakhutdinov (2008), we observe for RBMs that the bound provided by TRW(~µ; ~θ,ρ) over Lδ is loose and does not get better when optimizing over T.”
Deep Boltzmann Machines and the Centering Trick (2012, methodology): “Salakhutdinov (2008) introduced more elaborate procedures for particular types of Boltzmann machines such as restricted, semi-restricted and deep Boltzmann machines.”
Efficient Learning of Deep Boltzmann Machines (2010, unknown): “To alleviate this problem, (Salakhutdinov (2008); Salakhutdinov and Hinton (2009)) proposed a variational approach, where mean-field inference is used to estimate data-dependent expectations and an MCMC based stochastic approximation procedure is used to approximate the model’s expected sufficient…”
Learning in Markov Random Fields using Tempered Transitions (2009, methodology): “To estimate the models’ partition functions,we used Annealed Importance Sampling [10, 13] – a technique that is very similar to tempered transitions.”
Variational (Gradient) Estimate of the Score Function in Energy-based Latent Variable Models (2020, unknown): “The log-likelihood is commonly used to compare fully visible models on the same dataset and the annealed importance sampling (Neal, 2001; Salakhutdinov, 2008) is often used to estimate the partition function in the log-likelihood.”
Foundations and Advances in Deep Learning (2014, background): “Empirical evidence (Salakhutdinov, 2008) showed that the variance of the estimates is sufficiently small even with only a small number of runs.”
Learning Deep Boltzmann Machines using Adaptive MCMC (2010, methodology): “To estimate the models’ partition functions we used Annealed Importance Sampling (Neal, 2001; Salakhutdinov, 2008).”
Deep Boltzmann Machines (2009, methodology): “This is followed by applying SAP to update the model parameters θ (Salakhutdinov, 2008).”
A unified approach to building hybrid recommender systems (2009, methodology): “Investigating the use of recent advances in variational and Monte Carlo techniques for training Boltzmann machines [26] is another direction for future research.”