“An Abstraction-Refinement Approach to Verifying Convolutional Neural Networks” by Matan Ostrovsky, Clark Barrett, and Guy Katz. In Proceedings of the 20^th International Symposium on Automated Technology for Verification and Analysis (ATVA '22), (Ahmed Bouajjani, Lukás Holk, and Zhilin Wu, eds.), Oct. 2022, pp. 391-396.
Convolutional neural networks (CNNs) have achieved immense popularity in areas like computer vision, image processing, speech proccessing, and many others. Unfortunately, despite their excellent performance, they are prone to producing erroneous results — for example, minor perturbations to their inputs can result in severe classification errors. In this paper, we present the CNN-ABS framework, which implements an abstraction-refinement based scheme for CNN verification. Specifically, CNN-ABS simplifies the verification problem through the removal of convolutional connections in a way that soundly creates an over-approximation of the original problem; it then iteratively restores these connections if the resulting problem becomes too abstract. CNN-ABS is designed to use existing verification engines as a backend, and our evaluation demonstrates that it can significantly boost the performance of a state-of-the-art DNN verification engine, reducing runtime by 15.7% on average.
BibTeX entry:
@inproceedings{OBK22,
author = {Matan Ostrovsky and Clark Barrett and Guy Katz},
editor = {Ahmed Bouajjani and Luk{\'a}{\vs} Holk and Zhilin Wu},
title = {An Abstraction-Refinement Approach to Verifying Convolutional
Neural Networks},
booktitle = {Proceedings of the {\it 20^{th}} International Symposium
on Automated Technology for Verification and Analysis (ATVA '22)},
series = {Lecture Notes in Computer Science},
volume = {13505},
pages = {391--396},
publisher = {Springer International Publishing},
month = oct,
year = {2022},
doi = {10.1007/978-3-031-19992-9_25},
url = {http://theory.stanford.edu/~barrett/pubs/OBK22.pdf}
}
(This webpage was created with bibtex2web.)