Verifying Generalization in Deep Learning

Guy Amir*, Osher Maayan, Tom Zelazny, Guy Katz, Michael Schapira

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

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.

Original languageAmerican English
Title of host publicationComputer Aided Verification - 35th International Conference, CAV 2023, Proceedings
EditorsConstantin Enea, Akash Lal
PublisherSpringer Science and Business Media Deutschland GmbH
Pages438-455
Number of pages18
ISBN (Print)9783031377020
DOIs
StatePublished - 2023
EventProceedings of the 35th International Conference on Computer Aided Verification, CAV 2023 - Paris, France
Duration: 17 Jul 202322 Jul 2023

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13965 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

ConferenceProceedings of the 35th International Conference on Computer Aided Verification, CAV 2023
Country/TerritoryFrance
CityParis
Period17/07/2322/07/23

Bibliographical note

Publisher Copyright:
© 2023, The Author(s).

Fingerprint

Dive into the research topics of 'Verifying Generalization in Deep Learning'. Together they form a unique fingerprint.

Cite this