In this section
@inproceedings{Leofante:2023:10.1109/IJCNN54540.2023.10191169,
author = {Leofante, F and Henriksen, P and Lomuscio, A},
doi = {10.1109/IJCNN54540.2023.10191169},
pages = {1--9},
publisher = {IEEE},
title = {Verification-friendly networks: the case for parametric ReLUs},
url = {http://dx.doi.org/10.1109/IJCNN54540.2023.10191169},
year = {2023}
}
TY - CPAPER
AB - It has increasingly been recognised that verification can contribute to the validation and debugging of neural networks before deployment, particularly in safety-critical areas. While progress has been made in the area of verification of neural networks, present techniques still do not scale to large ReLU-based neural networks used in many applications. In this paper we show that considerable progress can be made by employing Parametric ReLU activation functions in lieu of plain ReLU functions. We give training procedures that produce networks which achieve one order of magnitude gain in verification overheads and 30-100% fewer timeouts with VeriNet, a SoA Symbolic Interval Propagation-based verification toolkit, while not compromising the resulting accuracy. Furthermore, we show that adversarial training combined with our approachimproves certified robustness up to 36% compared to adversarial training performed on baseline ReLU networks.
AU - Leofante,F
AU - Henriksen,P
AU - Lomuscio,A
DO - 10.1109/IJCNN54540.2023.10191169
EP - 9
PB - IEEE
PY - 2023///
SP - 1
TI - Verification-friendly networks: the case for parametric ReLUs
UR - http://dx.doi.org/10.1109/IJCNN54540.2023.10191169
UR - https://ieeexplore.ieee.org/document/10191169
UR - http://hdl.handle.net/10044/1/103943
ER -
Artificial Intelligence Network
South Kensington Campus
Imperial College London
SW7 2AZ
To reach the elected speaker of the network, Dr Rossella Arcucci, please contact:
To reach the network manager, Diana O'Malley - including to join the network - please contact: