Open Category Detection with PAC Guarantees

Abstract

Open category detection is the problem of detecting "alien" test instances that belong to categories or classes that were not present in the training data. In many applications, reliably detecting such aliens is central to ensuring the safety and accuracy of test set predictions. Unfortunately, there are no algorithms that provide theoretical guarantees on their ability to detect aliens under general assumptions. Further, while there are algorithms for open category detection, there are few empirical results that directly report alien detection rates. Thus, there are significant theoretical and empirical gaps in our understanding of open category detection. In this paper, we take a step toward addressing this gap by studying a simple, but practically-relevant variant of open category detection. In our setting, we are provided with a "clean" training set that contains only the target categories of interest and an unlabeled "contaminated" training set that contains a fraction $\alpha$ of alien examples. Under the assumption that we know an upper bound on $\alpha$, we develop an algorithm with PAC-style guarantees on the alien detection rate, while aiming to minimize false alarms. Empirical results on synthetic and standard benchmark datasets demonstrate the regimes in which the algorithm can be effective and provide a baseline for further advancements.

Cited by in corpus (15)

• Recent Advances in Open Set Recognition: A Survey
• Benchmarking Neural Network Robustness to Common Corruptions and Surface Variations
• Outlier Exposure with Confidence Control for Out-of-Distribution Detection
• Deep CNN-based Multi-task Learning for Open-Set Recognition
• Dense Uncertainty Estimation
• An Unbiased Risk Estimator for Learning with Augmented Classes
• Anomalous Example Detection in Deep Learning: A Survey
• Towards neural networks that provably know when they don't know
• Better the Devil you Know: An Analysis of Evasion Attacks using Out-of-Distribution Adversarial Examples
• DropCluster: A structured dropout for convolutional networks
• Understanding the Effect of Bias in Deep Anomaly Detection
• Resisting Out-of-Distribution Data Problem in Perturbation of XAI
• Bayesian OOD detection with aleatoric uncertainty and outlier exposure
• Utilizing Network Properties to Detect Erroneous Inputs
• Exploratory Machine Learning with Unknown Unknowns