On the Difficulty of Unbiased Alpha Divergence Minimization

Abstract

Several approximate inference algorithms have been proposed to minimize an alpha-divergence between an approximating distribution and a target distribution. Many of these algorithms introduce bias, the magnitude of which becomes problematic in high dimensions. Other algorithms are unbiased. These often seem to suffer from high variance, but little is rigorously known. In this work we study unbiased methods for alpha-divergence minimization through the Signal-to-Noise Ratio (SNR) of the gradient estimator. We study several representative scenarios where strong analytical results are possible, such as fully-factorized or Gaussian distributions. We find that when alpha is not zero, the SNR worsens exponentially in the dimensionality of the problem. This casts doubt on the practicality of these methods. We empirically confirm these theoretical results.

arXiv commentICML 2021

References in corpus (9)

• Adam: A Method for Stochastic Optimization
• Auto-Encoding Variational Bayes
• Sticking the Landing: Simple, Lower-Variance Gradient Estimators for Variational Inference
• Reweighted Wake-Sleep
• Advances in Variational Inference
• Alpha-Beta Divergence For Variational Inference
• Variational Inference with Tail-adaptive f-Divergence
• f-Divergence Variational Inference
• Markovian Score Climbing: Variational Inference with KL(p||q)

Cited by in corpus (3)

• User-friendly introduction to PAC-Bayes bounds
• Challenges and Opportunities in High-dimensional Variational Inference
• Mixture weights optimisation for Alpha-Divergence Variational Inference