Supervised Anomaly Detection based on Deep Autoregressive Density Estimators
This work addresses anomaly detection for domains where limited labeled anomaly data is available, offering a supervised approach that is incremental over existing density-based methods.
The paper tackles the problem of supervised anomaly detection by training a neural density estimator to maximize the likelihood of normal instances and minimize it for anomalous ones, using an autoregressive model for exact likelihood calculation. It demonstrates improved performance over existing methods on 16 datasets with a few labeled anomalies.
We propose a supervised anomaly detection method based on neural density estimators, where the negative log likelihood is used for the anomaly score. Density estimators have been widely used for unsupervised anomaly detection. By the recent advance of deep learning, the density estimation performance has been greatly improved. However, the neural density estimators cannot exploit anomaly label information, which would be valuable for improving the anomaly detection performance. The proposed method effectively utilizes the anomaly label information by training the neural density estimator so that the likelihood of normal instances is maximized and the likelihood of anomalous instances is lower than that of the normal instances. We employ an autoregressive model for the neural density estimator, which enables us to calculate the likelihood exactly. With the experiments using 16 datasets, we demonstrate that the proposed method improves the anomaly detection performance with a few labeled anomalous instances, and achieves better performance than existing unsupervised and supervised anomaly detection methods.