Iterative Refinement of the Approximate Posterior for Directed Belief Networks
نویسندگان
چکیده
Variational methods that rely on a recognition network to approximate the posterior of directed graphical models offer better inference and learning than previous methods. Recent advances that exploit the capacity and flexibility in this approach have expanded what kinds of models can be trained. However, as a proposal for the posterior, the capacity of the recognition network is limited, which can constrain the representational power of the generative model and increase the variance of Monte Carlo estimates. To address these issues, we introduce an iterative refinement procedure for improving the approximate posterior of the recognition network and show that training with the refined posterior is competitive with state-of-the-art methods. The advantages of refinement are further evident in an increased effective sample size, which implies a lower variance of gradient estimates.
منابع مشابه
Iterative Refinement of Approximate Posterior for Training Directed Belief Networks
Deep directed graphical models, while a potentially powerful class of generative representations, are challenging to train due to difficult inference. Recent advances in variational inference that make use of an inference or recognition network have advanced well beyond traditional variational inference and Markov chain Monte Carlo methods. While these techniques offer higher flexibility as wel...
متن کاملA Simple Insight into Iterative Belief Propagation's Success
In non-ergodic belief networks the posterior belief of many queries given evidence may become zero. The paper shows that when belief propagation is applied iteratively over arbitrary networks (the so called, iterative or loopy belief propagation (IBP)) it is identical to an arc-consistency algorithm relative to zero-belief queries (namely assessing zero posterior probabilities). This implies th...
متن کاملCategory: Algorithms and Architectures. Variational Belief Networks for Approximate Inference
Exact inference in large, densely connected probabilistic networks is computationally intractable, and approximate schemes are therefore of great importance. One approach is to use mean eld theory, in which the exact log-likelihood is bounded from below using a simpler approximating distribution. In the standard mean eld theory, the approximating distribution is factorial. We propose instead to...
متن کاملVariational Belief Networks for Approximate Inference
Exact inference in large, densely connected probabilistic networks is computa-tionally intractable, and approximate schemes are therefore of great importance. One approach is to use mean eld theory, in which the exact log-likelihood is bounded from below using a simpler approximating distribution. In the standard mean eld theory, the approximating distribution is factorial. We propose instead t...
متن کاملApproximation Algorithms for Probabilistic Reasoning: Sampling and Iterative Inference
The complexity of the exact inference increases exponentially with size and complexity of the network. As a result, the exact inference methods become impractical for large networks and we seek to approximate the results. A variety of approximation methods exist. This research focuses on two approximation methods for finding posterior marginals P (xi|e) in Bayesian networks: iterative belief up...
متن کامل