Extremely Simple Activation Shaping for Out-of-Distribution Detection

TL;DR

This paper introduces a simple, post-hoc activation shaping method called ASH that improves out-of-distribution detection in neural networks without additional training or data, achieving state-of-the-art results on ImageNet.

Contribution

The paper presents a novel, extremely simple activation shaping technique applied at inference time that enhances OOD detection without retraining or extra data.

Findings

ASH improves OOD detection performance on ImageNet.

ASH does not significantly reduce in-distribution accuracy.

The method is easy to implement and computationally efficient.

Abstract

The separation between training and deployment of machine learning models implies that not all scenarios encountered in deployment can be anticipated during training, and therefore relying solely on advancements in training has its limits. Out-of-distribution (OOD) detection is an important area that stress-tests a model's ability to handle unseen situations: Do models know when they don't know? Existing OOD detection methods either incur extra training steps, additional data or make nontrivial modifications to the trained network. In contrast, in this work, we propose an extremely simple, post-hoc, on-the-fly activation shaping method, ASH, where a large portion (e.g. 90%) of a sample's activation at a late layer is removed, and the rest (e.g. 10%) simplified or lightly adjusted. The shaping is applied at inference time, and does not require any statistics calculated from training data. Experiments show that such a simple treatment enhances in-distribution and out-of-distribution distinction so as to allow state-of-the-art OOD detection on ImageNet, and does not noticeably deteriorate the in-distribution accuracy. Video, animation and code can be found at: https://andrijazz.github.io/ash