Gaussian Latent Representations for Uncertainty Estimation using Mahalanobis Distance in Deep Classifiers

TL;DR

This paper introduces a lightweight, self-supervised regularization method that enhances Gaussian latent space representations for improved uncertainty estimation and out-of-distribution detection in deep classifiers, with minimal architectural changes.

Contribution

It proposes a novel, efficient regularization technique for Mahalanobis distance-based uncertainty prediction that automatically identifies and clusters non-Gaussian classes into Gaussian-like groups.

Findings

Achieves state-of-the-art OOD detection performance.

Maintains high predictive calibration with minimal inference overhead.

Effective in real-world microorganism classification tasks.

Abstract

Recent works show that the data distribution in a network's latent space is useful for estimating classification uncertainty and detecting Out-of-distribution (OOD) samples. To obtain a well-regularized latent space that is conducive for uncertainty estimation, existing methods bring in significant changes to model architectures and training procedures. In this paper, we present a lightweight, fast, and high-performance regularization method for Mahalanobis distance-based uncertainty prediction, and that requires minimal changes to the network's architecture. To derive Gaussian latent representation favourable for Mahalanobis Distance calculation, we introduce a self-supervised representation learning method that separates in-class representations into multiple Gaussians. Classes with non-Gaussian representations are automatically identified and dynamically clustered into multiple new classes that are approximately Gaussian. Evaluation on standard OOD benchmarks shows that our method achieves state-of-the-art results on OOD detection with minimal inference time, and is very competitive on predictive probability calibration. Finally, we show the applicability of our method to a real-life computer vision use case on microorganism classification.