TY - GEN
T1 - Minimal Multi-Layer Modifications of Deep Neural Networks
AU - Refaeli, Idan
AU - Katz, Guy
N1 - Publisher Copyright: © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2022
Y1 - 2022
N2 - Deep neural networks (DNNs) have become increasingly popular in recent years. However, despite their many successes, DNNs may also err and produce incorrect and potentially fatal outputs in safety-critical settings, such as autonomous driving, medical diagnosis, and airborne collision avoidance systems. Much work has been put into detecting such erroneous behavior in DNNs, e.g., via testing or verification, but removing these errors after their detection has received lesser attention. We present here a new tool, called 3M-DNN, for repairing a given DNN, which is known to err on some set of inputs. The novel repair procedure implemented in 3M-DNN computes a modification to the network’s weights that corrects its behavior, and attempts to minimize this change via a sequence of calls to a backend, black-box DNN verification engine. To the best of our knowledge, our method is the first one that allows repairing the network by simultaneously modifying multiple layers. This is achieved by splitting the network into sub-networks, and applying a single-layer repairing technique to each component. We evaluated 3M-DNN tool on an extensive set of benchmarks, obtaining promising results.
AB - Deep neural networks (DNNs) have become increasingly popular in recent years. However, despite their many successes, DNNs may also err and produce incorrect and potentially fatal outputs in safety-critical settings, such as autonomous driving, medical diagnosis, and airborne collision avoidance systems. Much work has been put into detecting such erroneous behavior in DNNs, e.g., via testing or verification, but removing these errors after their detection has received lesser attention. We present here a new tool, called 3M-DNN, for repairing a given DNN, which is known to err on some set of inputs. The novel repair procedure implemented in 3M-DNN computes a modification to the network’s weights that corrects its behavior, and attempts to minimize this change via a sequence of calls to a backend, black-box DNN verification engine. To the best of our knowledge, our method is the first one that allows repairing the network by simultaneously modifying multiple layers. This is achieved by splitting the network into sub-networks, and applying a single-layer repairing technique to each component. We evaluated 3M-DNN tool on an extensive set of benchmarks, obtaining promising results.
UR - http://www.scopus.com/inward/record.url?scp=85144822140&partnerID=8YFLogxK
U2 - https://doi.org/10.1007/978-3-031-21222-2_4
DO - https://doi.org/10.1007/978-3-031-21222-2_4
M3 - منشور من مؤتمر
SN - 9783031212215
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 46
EP - 66
BT - Software Verification and Formal Methods for ML-Enabled Autonomous Systems - 5th International Workshop, FoMLAS 2022, and 15th International Workshop, NSV 2022, Proceedings
A2 - Isac, Omri
A2 - Katz, Guy
A2 - Ivanov, Radoslav
A2 - Narodytska, Nina
A2 - Nenzi, Laura
PB - Springer Science and Business Media Deutschland GmbH
T2 - 5th International Workshop on Software Verification and Formal Methods for ML-Enables Autonomous Systems, FoMLAS 2022 and 15th International Workshop on Numerical Software Verification, NSV 2022
Y2 - 11 August 2022 through 11 August 2022
ER -