Mean-Shifted Contrastive Loss for Anomaly Detection

TL;DR

This paper introduces the Mean-Shifted Contrastive Loss, a novel approach that improves anomaly detection by effectively fine-tuning pre-trained representations, achieving state-of-the-art results on CIFAR-10.

Contribution

The paper proposes a modified contrastive loss function tailored for pre-trained features, enhancing anomaly detection performance.

Findings

Achieves 98.6% ROC-AUC on CIFAR-10

Demonstrates the effectiveness of the proposed loss in anomaly detection

Provides analysis of contrastive learning challenges with pre-trained features

Abstract

Deep anomaly detection methods learn representations that separate between normal and anomalous images. Although self-supervised representation learning is commonly used, small dataset sizes limit its effectiveness. It was previously shown that utilizing external, generic datasets (e.g. ImageNet classification) can significantly improve anomaly detection performance. One approach is outlier exposure, which fails when the external datasets do not resemble the anomalies. We take the approach of transferring representations pre-trained on external datasets for anomaly detection. Anomaly detection performance can be significantly improved by fine-tuning the pre-trained representations on the normal training images. In this paper, we first demonstrate and analyze that contrastive learning, the most popular self-supervised learning paradigm cannot be naively applied to pre-trained features. The reason is that pre-trained feature initialization causes poor conditioning for standard contrastive objectives, resulting in bad optimization dynamics. Based on our analysis, we provide a modified contrastive objective, the Mean-Shifted Contrastive Loss. Our method is highly effective and achieves a new state-of-the-art anomaly detection performance including $98.6\%$ ROC-AUC on the CIFAR-10 dataset.