Partition Functions from Rao-Blackwellized Tempered Sampling
arXiv:1603.01912
Abstract
Partition functions of probability distributions are important quantities for model evaluation and comparisons. We present a new method to compute partition functions of complex and multimodal distributions. Such distributions are often sampled using simulated tempering, which augments the target space with an auxiliary inverse temperature variable. Our method exploits the multinomial probability law of the inverse temperatures, and provides estimates of the partition function in terms of a simple quotient of Rao-Blackwellized marginal inverse temperature probability estimates, which are updated while sampling. We show that the method has interesting connections with several alternative popular methods, and offers some significant advantages. In particular, we empirically find that the new method provides more accurate estimates than Annealed Importance Sampling when calculating partition functions of large Restricted Boltzmann Machines (RBM); moreover, the method is sufficiently accurate to track training and validation log-likelihoods during learning of RBMs, at minimal computational cost.
15 pages, 8 figures; Appearing at International Conference on Machine Learning 2016
References in corpus (9)
- MCMC using Hamiltonian dynamics
- An efficient, multiple range random walk algorithm to calculate the density of states
- Statistically optimal analysis of samples from multiple equilibrium states
- Simulated Tempering: A New Monte Carlo Scheme
- Importance sampling methods for Bayesian discrimination between embedded models
- Estimating Ratios of Normalizing Constants Using Linked Importance Sampling
- Minimum Probability Flow Learning
- Sandwiching the marginal likelihood using bidirectional Monte Carlo
- Understanding Minimum Probability Flow for RBMs Under Various Kinds of Dynamics