Verification-Aided Deep Ensemble Selection

Guy Amir, Tom Zelazny, Guy Katz, Michael Schapira

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

12 Scopus citations

Abstract

Deep neural networks (DNNs) have become the technology of choice for realizing a variety of complex tasks. However, as highlighted by many recent studies, even an imperceptible perturbation to a correctly classified input can lead to misclassification by a DNN. This renders DNNs vulnerable to strategic input manipulations by attackers, and also over-sensitive to environmental noise. To mitigate this phenomenon, practitioners apply joint classification by an ensemble of DNNs. By aggregating the classification outputs of different individual DNNs for the same input, ensemble-based classification reduces the risk of misclassifications due to the specific realization of the stochastic training process of any single DNN. However, the effectiveness of a DNN ensemble is highly dependent on its members not simultaneously erring on many different inputs. In this case study, we harness recent advances in DNN verification to devise a methodology for identifying ensemble compositions that are less prone to simultaneous errors, even when the input is adversarially perturbed - resulting in more robustly-accurate ensemble-based classification. Our proposed framework uses a DNN verifier as a backend, and includes heuristics that help reduce the high complexity of directly verifying ensembles. More broadly, our work puts forth a novel universal objective for formal verification that can potentially improve the robustness of real-world, deep-learning-based systems across a variety of application domains.

Original languageEnglish
Title of host publicationProceedings of the 22nd Conference on Formal Methods in Computer-Aided Design, FMCAD 2022
EditorsAlberto Griggio, Neha Rungta
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages27-37
Number of pages11
ISBN (Electronic)9783854480532
DOIs
StatePublished - 2022
Event22nd International Conference on Formal Methods in Computer-Aided Design, FMCAD 2022 - Trento, Italy
Duration: 17 Oct 202221 Oct 2022

Publication series

NameProceedings of the 22nd Conference on Formal Methods in Computer-Aided Design, FMCAD 2022

Conference

Conference22nd International Conference on Formal Methods in Computer-Aided Design, FMCAD 2022
Country/TerritoryItaly
CityTrento
Period17/10/2221/10/22

Bibliographical note

Publisher Copyright:
© 2022 FMCAD Association and authors.

Fingerprint

Dive into the research topics of 'Verification-Aided Deep Ensemble Selection'. Together they form a unique fingerprint.

Cite this