Quantile-based Maximum Likelihood Training for Outlier Detection

TL;DR

This paper proposes a novel quantile-based maximum likelihood training method using normalizing flows on discriminative features to improve outlier detection, reducing reliance on negative training data and outperforming existing unsupervised approaches.

Contribution

It introduces a quantile-based maximum likelihood objective with normalizing flows for inlier distribution modeling, enhancing outlier detection without extensive negative data.

Findings

Outperforms state-of-the-art unsupervised outlier detection methods

Competitive with recent self-supervised approaches

Reduces dependency on negative training data

Abstract

Discriminative learning effectively predicts true object class for image classification. However, it often results in false positives for outliers, posing critical concerns in applications like autonomous driving and video surveillance systems. Previous attempts to address this challenge involved training image classifiers through contrastive learning using actual outlier data or synthesizing outliers for self-supervised learning. Furthermore, unsupervised generative modeling of inliers in pixel space has shown limited success for outlier detection. In this work, we introduce a quantile-based maximum likelihood objective for learning the inlier distribution to improve the outlier separation during inference. Our approach fits a normalizing flow to pre-trained discriminative features and detects the outliers according to the evaluated log-likelihood. The experimental evaluation demonstrates the effectiveness of our method as it surpasses the performance of the state-of-the-art unsupervised methods for outlier detection. The results are also competitive compared with a recent self-supervised approach for outlier detection. Our work allows to reduce dependency on well-sampled negative training data, which is especially important for domains like medical diagnostics or remote sensing.