Detecting Out-of-Distribution Examples with In-distribution Examples and Gram Matrices

TL;DR

This paper introduces a novel OOD detection method using Gram matrices to identify inconsistencies in activity patterns of neural networks, which works with pre-trained models and outperforms existing methods.

Contribution

The proposed approach detects OOD examples by analyzing Gram matrix anomalies without needing OOD data for tuning, applicable across various models and datasets.

Findings

High OOD detection accuracy across multiple datasets.

Outperforms or matches state-of-the-art methods.

Does not require OOD data for hyperparameter tuning.

Abstract

When presented with Out-of-Distribution (OOD) examples, deep neural networks yield confident, incorrect predictions. Detecting OOD examples is challenging, and the potential risks are high. In this paper, we propose to detect OOD examples by identifying inconsistencies between activity patterns and class predicted. We find that characterizing activity patterns by Gram matrices and identifying anomalies in gram matrix values can yield high OOD detection rates. We identify anomalies in the gram matrices by simply comparing each value with its respective range observed over the training data. Unlike many approaches, this can be used with any pre-trained softmax classifier and does not require access to OOD data for fine-tuning hyperparameters, nor does it require OOD access for inferring parameters. The method is applicable across a variety of architectures and vision datasets and, for the important and surprisingly hard task of detecting far-from-distribution out-of-distribution examples, it generally performs better than or equal to state-of-the-art OOD detection methods (including those that do assume access to OOD examples).