Improving Predictive Uncertainty Estimation using Dropout -- Hamiltonian Monte Carlo

TL;DR

This paper introduces a novel method combining Dropout and Hamiltonian Monte Carlo to improve predictive uncertainty estimation in classification tasks, addressing challenges with non-smooth energy functions and enhancing generalization.

Contribution

It presents a robust methodology that integrates Dropout with HMC, overcoming convergence issues and improving uncertainty estimation in Bayesian inference for neural networks.

Findings

Empirical success in predictive uncertainty estimation

Improved generalization on difficult test examples

Effective handling of non-smooth energy functions

Abstract

Estimating predictive uncertainty is crucial for many computer vision tasks, from image classification to autonomous driving systems. Hamiltonian Monte Carlo (HMC) is an sampling method for performing Bayesian inference. On the other hand, Dropout regularization has been proposed as an approximate model averaging technique that tends to improve generalization in large scale models such as deep neural networks. Although, HMC provides convergence guarantees for most standard Bayesian models, it does not handle discrete parameters arising from Dropout regularization. In this paper, we present a robust methodology for improving predictive uncertainty in classification problems, based on Dropout and Hamiltonian Monte Carlo. Even though Dropout induces a non-smooth energy function with no such convergence guarantees, the resulting discretization of the Hamiltonian proves empirical success. The proposed method allows to effectively estimate the predictive accuracy and to provide better generalization for difficult test examples.