Towards Explaining Anomalies: A Deep Taylor Decomposition of One-Class Models
This addresses the need for interpretability in anomaly detection for practitioners, though it is incremental as it builds on existing deep Taylor decomposition techniques.
The paper tackles the problem of explaining anomaly predictions in one-class SVMs by introducing a method that decomposes outlier predictions into input variable contributions, achieving reliable explanations and outperforming baselines like sensitivity analysis.
A common machine learning task is to discriminate between normal and anomalous data points. In practice, it is not always sufficient to reach high accuracy at this task, one also would like to understand why a given data point has been predicted in a certain way. We present a new principled approach for one-class SVMs that decomposes outlier predictions in terms of input variables. The method first recomposes the one-class model as a neural network with distance functions and min-pooling, and then performs a deep Taylor decomposition (DTD) of the model output. The proposed One-Class DTD is applicable to a number of common distance-based SVM kernels and is able to reliably explain a wide set of data anomalies. Furthermore, it outperforms baselines such as sensitivity analysis, nearest neighbor, or simple edge detection.