Verifying Generalization in Deep Learning

TL;DR

This paper introduces a verification-based method to assess and improve the generalization of deep neural networks across different input domains, addressing a key challenge in deploying DNNs in real-world, variable environments.

Contribution

It proposes a novel verification-driven approach that measures DNN generalization by agreement among independently trained models, demonstrated on reinforcement learning benchmarks.

Findings

Effective in identifying well-generalizing DNNs

Applicable to reinforcement learning and real-world systems

Establishes a new formal verification objective

Abstract

Deep neural networks (DNNs) are the workhorses of deep learning, which constitutes the state of the art in numerous application domains. However, DNN-based decision rules are notoriously prone to poor generalization, i.e., may prove inadequate on inputs not encountered during training. This limitation poses a significant obstacle to employing deep learning for mission-critical tasks, and also in real-world environments that exhibit high variability. We propose a novel, verification-driven methodology for identifying DNN-based decision rules that generalize well to new input domains. Our approach quantifies generalization to an input domain by the extent to which decisions reached by independently trained DNNs are in agreement for inputs in this domain. We show how, by harnessing the power of DNN verification, our approach can be efficiently and effectively realized. We evaluate our verification-based approach on three deep reinforcement learning (DRL) benchmarks, including a system for Internet congestion control. Our results establish the usefulness of our approach. More broadly, our work puts forth a novel objective for formal verification, with the potential for mitigating the risks associated with deploying DNN-based systems in the wild.