Positive Difference Distribution for Image Outlier Detection using Normalizing Flows and Contrastive Data
This addresses the need for more reliable outlier detection in machine learning, particularly for safety-critical applications, though it appears incremental as it builds on existing normalizing flow and contrastive learning techniques.
The paper tackled the problem of poor outlier detection performance using standard generative model likelihoods by proposing a method that trains a normalizing flow with a contrastive dataset to learn a probabilistic outlier score, achieving improved results on benchmark datasets compared to existing methods.
Detecting test data deviating from training data is a central problem for safe and robust machine learning. Likelihoods learned by a generative model, e.g., a normalizing flow via standard log-likelihood training, perform poorly as an outlier score. We propose to use an unlabelled auxiliary dataset and a probabilistic outlier score for outlier detection. We use a self-supervised feature extractor trained on the auxiliary dataset and train a normalizing flow on the extracted features by maximizing the likelihood on in-distribution data and minimizing the likelihood on the contrastive dataset. We show that this is equivalent to learning the normalized positive difference between the in-distribution and the contrastive feature density. We conduct experiments on benchmark datasets and compare to the likelihood, the likelihood ratio and state-of-the-art anomaly detection methods.