A Simple and Effective Method for Uncertainty Quantification and OOD Detection

TL;DR

This paper introduces a computationally efficient method for uncertainty quantification and out-of-distribution detection using feature space density, outperforming existing Bayesian and ensemble approaches.

Contribution

The authors propose a novel approach leveraging feature space density via kernel density estimation for effective uncertainty quantification and OOD detection with a single deterministic model.

Findings

Outperforms baseline models in synthetic and real-world datasets

Effectively detects distributional shifts and OOD samples

Uses a simple, scalable method based on feature space density

Abstract

Bayesian neural networks and deep ensemble methods have been proposed for uncertainty quantification; however, they are computationally intensive and require large storage. By utilizing a single deterministic model, we can solve the above issue. We propose an effective method based on feature space density to quantify uncertainty for distributional shifts and out-of-distribution (OOD) detection. Specifically, we leverage the information potential field derived from kernel density estimation to approximate the feature space density of the training set. By comparing this density with the feature space representation of test samples, we can effectively determine whether a distributional shift has occurred. Experiments were conducted on a 2D synthetic dataset (Two Moons and Three Spirals) as well as an OOD detection task (CIFAR-10 vs. SVHN). The results demonstrate that our method outperforms baseline models.