Towards Optimal Feature-Shaping Methods for Out-of-Distribution Detection

TL;DR

This paper introduces a unified optimization framework for feature-shaping methods in out-of-distribution detection, proposing a new approach that enhances generalization across diverse datasets and models.

Contribution

It formulates an abstract optimization framework, simplifies it with a piecewise constant function, and provides a closed-form solution using only in-distribution data.

Findings

Improved OOD detection performance across various datasets.

Existing methods approximate the optimal solution within the framework.

The proposed method generalizes better to unseen OOD data.

Abstract

Feature shaping refers to a family of methods that exhibit state-of-the-art performance for out-of-distribution (OOD) detection. These approaches manipulate the feature representation, typically from the penultimate layer of a pre-trained deep learning model, so as to better differentiate between in-distribution (ID) and OOD samples. However, existing feature-shaping methods usually employ rules manually designed for specific model architectures and OOD datasets, which consequently limit their generalization ability. To address this gap, we first formulate an abstract optimization framework for studying feature-shaping methods. We then propose a concrete reduction of the framework with a simple piecewise constant shaping function and show that existing feature-shaping methods approximate the optimal solution to the concrete optimization problem. Further, assuming that OOD data is inaccessible, we propose a formulation that yields a closed-form solution for the piecewise constant shaping function, utilizing solely the ID data. Through extensive experiments, we show that the feature-shaping function optimized by our method improves the generalization ability of OOD detection across a large variety of datasets and model architectures.