A Survey on Assessing the Generalization Envelope of Deep Neural Networks: Predictive Uncertainty, Out-of-distribution and Adversarial Samples

TL;DR

This survey reviews methods for assessing whether deep neural networks will generalize correctly to specific inputs, focusing on predictive uncertainty, out-of-distribution detection, and adversarial sample identification to improve reliability.

Contribution

It unifies three research areas under the framework of generalization assessment and provides a structured overview of inference-time methods for DNN reliability evaluation.

Findings

Identifies commonalities among uncertainty, OOD, and adversarial detection methods.

Highlights promising approaches for real-time generalization assessment.

Provides a comprehensive overview of current techniques in the field.

Abstract

Deep Neural Networks (DNNs) achieve state-of-the-art performance on numerous applications. However, it is difficult to tell beforehand if a DNN receiving an input will deliver the correct output since their decision criteria are usually nontransparent. A DNN delivers the correct output if the input is within the area enclosed by its generalization envelope. In this case, the information contained in the input sample is processed reasonably by the network. It is of large practical importance to assess at inference time if a DNN generalizes correctly. Currently, the approaches to achieve this goal are investigated in different problem set-ups rather independently from one another, leading to three main research and literature fields: predictive uncertainty, out-of-distribution detection and adversarial example detection. This survey connects the three fields within the larger framework of investigating the generalization performance of machine learning methods and in particular DNNs. We underline the common ground, point at the most promising approaches and give a structured overview of the methods that provide at inference time means to establish if the current input is within the generalization envelope of a DNN.