Anomaly scores for generative models
This work addresses anomaly detection for data modeled by generative models, such as auto-encoders, but is incremental as it builds on existing methods to improve theoretical justification.
The paper tackled the problem of anomaly detection using generative models by proposing a new anomaly score that is theoretically compatible, addressing the ill-defined nature of reconstruction error which assumes anomalies lie outside a learned manifold. The result includes a method for selecting hyper-parameters and models, and explains why reconstruction error performs well experimentally despite theoretical weaknesses.
Reconstruction error is a prevalent score used to identify anomalous samples when data are modeled by generative models, such as (variational) auto-encoders or generative adversarial networks. This score relies on the assumption that normal samples are located on a manifold and all anomalous samples are located outside. Since the manifold can be learned only where the training data lie, there are no guarantees how the reconstruction error behaves elsewhere and the score, therefore, seems to be ill-defined. This work defines an anomaly score that is theoretically compatible with generative models, and very natural for (variational) auto-encoders as they seem to be prevalent. The new score can be also used to select hyper-parameters and models. Finally, we explain why reconstruction error delivers good experimental results despite weak theoretical justification.