pyCLAD: The universal framework for continual lifelong anomaly detection

Abstract

Anomaly detection is a recognized problem with high significance and impact in many real-world settings. Continual anomaly detection is an emerging paradigm that allows for the design of anomaly detection methods capable of adapting to new challenges in dynamic environments while retaining past knowledge. In this paper, we propose pyCLAD, the first software framework providing foundations for the design of new continual anomaly detection scenarios, strategies, and evaluation protocols, while streamlining the execution of experimental workflows with high reproducibility standards.

Keywords

• Continual anomaly detection
• Lifelong anomaly detection
• Continual learning
• Anomaly detection
• Software