Variational Bayesian Decision-making for Continuous Utilities

Abstract

Bayesian decision theory outlines a rigorous framework for making optimal decisions based on maximizing expected utility over a model posterior. However, practitioners often do not have access to the full posterior and resort to approximate inference strategies. In such cases, taking the eventual decision-making task into account while performing the inference allows for calibrating the posterior approximation to maximize the utility. We present an automatic pipeline that co-opts continuous utilities into variational inference algorithms to account for decision-making. We provide practical strategies for approximating and maximizing the gain, and empirically demonstrate consistent improvement when calibrating approximations for specific utilities.

Appearing at Neural Information Processing Systems 32 (NeurIPS 2019)

References in corpus (9)

• Expectation Propagation for approximate Bayesian inference
• Edward: A library for probabilistic modeling, inference, and criticism
• Implicit Reparameterization Gradients
• Yes, but Did It Work?: Evaluating Variational Inference
• Generalized Variational Inference: Three arguments for deriving new Posteriors
• Boosting Variational Inference
• Loss-Calibrated Approximate Inference in Bayesian Neural Networks
• Adaptive learning rates and parallelization for stochastic, sparse, non-smooth gradients
• Boosting Black Box Variational Inference