TY - GEN
T1 - Verifying Generalization in Deep Learning
AU - Amir, Guy
AU - Maayan, Osher
AU - Zelazny, Tom
AU - Katz, Guy
AU - Schapira, Michael
N1 - Publisher Copyright: © 2023, The Author(s).
PY - 2023
Y1 - 2023
N2 - 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.
AB - 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.
UR - http://www.scopus.com/inward/record.url?scp=85172280070&partnerID=8YFLogxK
U2 - https://doi.org/10.1007/978-3-031-37703-7_21
DO - https://doi.org/10.1007/978-3-031-37703-7_21
M3 - منشور من مؤتمر
SN - 9783031377020
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 438
EP - 455
BT - Computer Aided Verification - 35th International Conference, CAV 2023, Proceedings
A2 - Enea, Constantin
A2 - Lal, Akash
PB - Springer Science and Business Media Deutschland GmbH
T2 - Proceedings of the 35th International Conference on Computer Aided Verification, CAV 2023
Y2 - 17 July 2023 through 22 July 2023
ER -