Uncertainty-Aware Out-of-Distribution Detection with Gaussian Processes

2024-12-30

AI Summary

To address the challenge of reliably detecting out-of-distribution (OOD) inputs for deep neural networks (DNNs) in open-world settings, this paper proposes an OOD detection method that relies solely on in-distribution (InD) training data. The core innovation is the first integration of multi-class Gaussian processes into the DNN’s softmax output layer, enabling explicit modeling of predictive uncertainty. Leveraging the posterior predictive distribution, our approach adaptively learns an InD/OOD decision boundary without requiring any OOD samples during training. Evaluated on standard and real-world image benchmarks, the method achieves state-of-the-art performance: under the zero-OOD-sample training setting, it improves OOD detection AUC by 5.2–8.7 percentage points over prior approaches. This advancement significantly enhances model robustness and safety when encountering previously unseen data.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Computer Vision: Object Detection & Categorization

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Deep neural networks (DNNs) are often constructed under the closed-world assumption, which may fail to generalize to the out-of-distribution (OOD) data. This leads to DNNs producing overconfident wrong predictions and can result in disastrous consequences in safety-critical applications. Existing OOD detection methods mainly rely on curating a set of OOD data for model training or hyper-parameter tuning to distinguish OOD data from training data (also known as in-distribution data or InD data). However, OOD samples are not always available during the training phase in real-world applications, hindering the OOD detection accuracy. To overcome this limitation, we propose a Gaussian-process-based OOD detection method to establish a decision boundary based on InD data only. The basic idea is to perform uncertainty quantification of the unconstrained softmax scores of a DNN via a multi-class Gaussian process (GP), and then define a score function to separate InD and potential OOD data based on their fundamental differences in the posterior predictive distribution from the GP. Two case studies on conventional image classification datasets and real-world image datasets are conducted to demonstrate that the proposed method outperforms the state-of-the-art OOD detection methods when OOD samples are not observed in the training phase.

Problem

• Unseen Data Handling
• Deep Neural Networks
• Anomaly Detection

Innovation

• Gaussian Processes
• Deep Neural Networks Uncertainty Estimation
• Anomaly Detection in Image Recognition