Meta-Learning MCMC Proposals

Abstract

Effective implementations of sampling-based probabilistic inference often require manually constructed, model-specific proposals. Inspired by recent progresses in meta-learning for training learning agents that can generalize to unseen environments, we propose a meta-learning approach to building effective and generalizable MCMC proposals. We parametrize the proposal as a neural network to provide fast approximations to block Gibbs conditionals. The learned neural proposals generalize to occurrences of common structural motifs across different models, allowing for the construction of a library of learned inference primitives that can accelerate inference on unseen models with no model-specific training required. We explore several applications including open-universe Gaussian mixture models, in which our learned proposals outperform a hand-tuned sampler, and a real-world named entity recognition task, in which our sampler yields higher final F1 scores than classical single-site Gibbs sampling.

32nd Conference on Neural Information Processing Systems (NeurIPS 2018), Montreal, Canada

Cited by in corpus (11)

• Neural Relational Inference with Fast Modular Meta-learning
• Markovian Score Climbing: Variational Inference with KL(p||q)
• Solving Compositional Reinforcement Learning Problems via Task Reduction
• Deep Involutive Generative Models for Neural MCMC
• Learning Proposals for Probabilistic Programs with Inference Combinators
• Approximate Inference in Discrete Distributions with Monte Carlo Tree Search and Value Functions
• Amortized Population Gibbs Samplers with Neural Sufficient Statistics
• Meta-Learning an Inference Algorithm for Probabilistic Programs
• Is High Variance Unavoidable in RL? A Case Study in Continuous Control
• Reparameterized Sampling for Generative Adversarial Networks
• LSB: Local Self-Balancing MCMC in Discrete Spaces