Post-hoc loss-calibration for Bayesian neural networks
This work addresses the challenge of decision-making under uncertainty for users of Bayesian neural networks, offering an incremental improvement over existing loss-calibrated techniques.
The paper tackles the problem of making high-utility decisions with approximate Bayesian neural networks by developing a post-hoc loss-calibration method that corrects posterior predictive distributions, resulting in empirically higher quality decisions across diverse tasks and datasets.
Bayesian decision theory provides an elegant framework for acting optimally under uncertainty when tractable posterior distributions are available. Modern Bayesian models, however, typically involve intractable posteriors that are approximated with, potentially crude, surrogates. This difficulty has engendered loss-calibrated techniques that aim to learn posterior approximations that favor high-utility decisions. In this paper, focusing on Bayesian neural networks, we develop methods for correcting approximate posterior predictive distributions encouraging them to prefer high-utility decisions. In contrast to previous work, our approach is agnostic to the choice of the approximate inference algorithm, allows for efficient test time decision making through amortization, and empirically produces higher quality decisions. We demonstrate the effectiveness of our approach through controlled experiments spanning a diversity of tasks and datasets.