Elsa: Energy-based learning for semi-supervised anomaly detection

TL;DR

Elsa is a semi-supervised anomaly detection method combining energy-based models and contrastive learning, designed to be robust against contaminated training data, achieving state-of-the-art results across various scenarios.

Contribution

Elsa introduces a novel energy-based semi-supervised approach that enhances anomaly detection robustness in contaminated data environments, with a new energy function and theoretical insights.

Findings

Elsa outperforms existing methods in contaminated scenarios.

The energy-based fine-tuning improves anomaly detection accuracy.

Contrastive learning alone is insufficient under data contamination.

Abstract

Anomaly detection aims at identifying deviant instances from the normal data distribution. Many advances have been made in the field, including the innovative use of unsupervised contrastive learning. However, existing methods generally assume clean training data and are limited when the data contain unknown anomalies. This paper presents Elsa, a novel semi-supervised anomaly detection approach that unifies the concept of energy-based models with unsupervised contrastive learning. Elsa instills robustness against any data contamination by a carefully designed fine-tuning step based on the new energy function that forces the normal data to be divided into classes of prototypes. Experiments on multiple contamination scenarios show the proposed model achieves SOTA performance. Extensive analyses also verify the contribution of each component in the proposed model. Beyond the experiments, we also offer a theoretical interpretation of why contrastive learning alone cannot detect anomalies under data contamination.