Uncertainty in the Variational Information Bottleneck

TL;DR

This paper demonstrates that the Variational Information Bottleneck (VIB) can enhance a neural network's calibration and out-of-distribution detection capabilities by providing natural uncertainty metrics without explicit design for these tasks.

Contribution

The study shows that VIB inherently improves uncertainty quantification and out-of-distribution detection in neural networks, offering new insights into its benefits.

Findings

VIB improves classification calibration.

VIB enhances out-of-distribution detection.

VIB provides natural uncertainty metrics.

Abstract

We present a simple case study, demonstrating that Variational Information Bottleneck (VIB) can improve a network's classification calibration as well as its ability to detect out-of-distribution data. Without explicitly being designed to do so, VIB gives two natural metrics for handling and quantifying uncertainty.