UAI 2021poster50 citations
Formal verification of neural networks for safety-critical tasks in deep reinforcement learning
Davide Corsi, Enrico Marchesini, Alessandro Farinelli
Abstract
In the last years, neural networks achieved groundbreaking successes in a wide variety of applications. However, for safety critical tasks, such as robotics and healthcare, it is necessary to provide some specific guarantees before the deployment in a real world context. Even in these scenarios, where high cost equipment and human safety are involved, the evaluation of the models is usually performed with the standard metrics (i.e., cumulative reward or success rate). In this paper, we introduce a novel metric for the evaluation of models in safety critical tasks, the
BibTeX
@InProceedings{pmlr-v161-corsi21a,
title = {Formal verification of neural networks for safety-critical tasks in deep reinforcement learning},
author = {Corsi, Davide and Marchesini, Enrico and Farinelli, Alessandro},
booktitle = {Proceedings of the Thirty-Seventh Conference on Uncertainty in Artificial Intelligence},
pages = {333--343},
year = {2021},
editor = {de Campos, Cassio and Maathuis, Marloes H.},
volume = {161},
series = {Proceedings of Machine Learning Research},
month = {27--30 Jul},
publisher = {PMLR},
pdf = {https://proceedings.mlr.press/v161/corsi21a/corsi21a.pdf},
url = {https://proceedings.mlr.press/v161/corsi21a.html},
abstract = {In the last years, neural networks achieved groundbreaking successes in a wide variety of applications. However, for safety critical tasks, such as robotics and healthcare, it is necessary to provide some specific guarantees before the deployment in a real world context. Even in these scenarios, where high cost equipment and human safety are involved, the evaluation of the models is usually performed with the standard metrics (i.e., cumulative reward or success rate). In this paper, we introduce a novel metric for the evaluation of models in safety critical tasks, the