Knowledge-Augmented Explainable and Interpretable Learning for Anomaly Detection and Diagnosis
Martin Atzmueller, Tim Bohne, Patricia Windler
TL;DR
Knowledge-Augmented Explainable and Interpretable Learning for Anomaly Detection and Diagnosis addresses the need for transparent AI systems in anomaly detection across technical domains. It surveys three strands: knowledge-augmented pattern mining (notably subgroup discovery), learning and refining diagnostic scoring systems, and neuro-symbolic approaches that fuse knowledge graphs with neural models for explainable diagnosis. Concrete applications include an industrial logistics anomaly-detection use case and an automotive neuro-symbolic fault-diagnosis framework built on ontologies, SPARQL tooling, and time-series oscillogram classification. The chapter emphasizes how explicit knowledge representations, interactive refinement, and causal reasoning via knowledge graphs enhance interpretability, trust, and computational sensemaking, and highlights future work toward more integrated and generalizable neuro-symbolic architectures.
Abstract
Knowledge-augmented learning enables the combination of knowledge-based and data-driven approaches. For anomaly detection and diagnosis, understandability is typically an important factor, especially in high-risk areas. Therefore, explainability and interpretability are also major criteria in such contexts. This chapter focuses on knowledge-augmented explainable and interpretable learning to enhance understandability, transparency and ultimately computational sensemaking. We exemplify different approaches and methods in the domains of anomaly detection and diagnosis - from comparatively simple interpretable methods towards more advanced neuro-symbolic approaches.