{"year":"2021","language":[{"iso":"eng"}],"day":"01","date_updated":"2023-06-23T07:01:11Z","main_file_link":[{"open_access":"1","url":"https://proceedings.neurips.cc/paper/2021/hash/544defa9fddff50c53b71c43e0da72be-Abstract.html"}],"conference":{"end_date":"2021-12-10","name":"NeurIPS: Neural Information Processing Systems","location":"Virtual","start_date":"2021-12-06"},"publication":"35th Conference on Neural Information Processing Systems","ec_funded":1,"doi":"10.48550/arXiv.2111.03165","oa_version":"Published Version","tmp":{"image":"/images/cc_by_nc_nd.png","short":"CC BY-NC-ND (3.0)","name":"Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported (CC BY-NC-ND 3.0)","legal_code_url":"https://creativecommons.org/licenses/by-nc-nd/3.0/legalcode"},"license":"https://creativecommons.org/licenses/by-nc-nd/3.0/","abstract":[{"lang":"eng","text":"Bayesian neural networks (BNNs) place distributions over the weights of a neural network to model uncertainty in the data and the network's prediction. We consider the problem of verifying safety when running a Bayesian neural network policy in a feedback loop with infinite time horizon systems. Compared to the existing sampling-based approaches, which are inapplicable to the infinite time horizon setting, we train a separate deterministic neural network that serves as an infinite time horizon safety certificate. In particular, we show that the certificate network guarantees the safety of the system over a subset of the BNN weight posterior's support. Our method first computes a safe weight set and then alters the BNN's weight posterior to reject samples outside this set. Moreover, we show how to extend our approach to a safe-exploration reinforcement learning setting, in order to avoid unsafe trajectories during the training of the policy. We evaluate our approach on a series of reinforcement learning benchmarks, including non-Lyapunovian safety specifications."}],"user_id":"2EBD1598-F248-11E8-B48F-1D18A9856A87","oa":1,"external_id":{"arxiv":["2111.03165"]},"has_accepted_license":"1","file":[{"file_size":452492,"relation":"main_file","date_created":"2022-01-26T07:39:59Z","file_id":"10682","date_updated":"2022-01-26T07:39:59Z","content_type":"application/pdf","success":1,"creator":"mlechner","file_name":"infinite_time_horizon_safety_o.pdf","access_level":"open_access","checksum":"0fc0f852525c10dda9cc9ffea07fb4e4"}],"_id":"10667","status":"public","quality_controlled":"1","publication_status":"published","department":[{"_id":"GradSch"},{"_id":"ToHe"},{"_id":"KrCh"}],"article_processing_charge":"No","ddc":["000"],"type":"conference","alternative_title":[" Advances in Neural Information Processing Systems"],"date_published":"2021-12-01T00:00:00Z","title":"Infinite time horizon safety of Bayesian neural networks","project":[{"name":"International IST Doctoral Program","grant_number":"665385","call_identifier":"H2020","_id":"2564DBCA-B435-11E9-9278-68D0E5697425"},{"name":"Formal Methods for Stochastic Models: Algorithms and Applications","call_identifier":"H2020","grant_number":"863818","_id":"0599E47C-7A3F-11EA-A408-12923DDC885E"},{"_id":"25F42A32-B435-11E9-9278-68D0E5697425","name":"The Wittgenstein Prize","call_identifier":"FWF","grant_number":"Z211"}],"file_date_updated":"2022-01-26T07:39:59Z","acknowledgement":"This research was supported in part by the Austrian Science Fund (FWF) under grant Z211-N23 (Wittgenstein Award), ERC CoG 863818 (FoRM-SMArt), and the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie Grant Agreement No. 665385.","date_created":"2022-01-25T15:45:58Z","related_material":{"record":[{"id":"11362","status":"public","relation":"dissertation_contains"}]},"citation":{"short":"M. Lechner, Ð. Žikelić, K. Chatterjee, T.A. Henzinger, in:, 35th Conference on Neural Information Processing Systems, 2021.","chicago":"Lechner, Mathias, Ðorđe Žikelić, Krishnendu Chatterjee, and Thomas A Henzinger. “Infinite Time Horizon Safety of Bayesian Neural Networks.” In 35th Conference on Neural Information Processing Systems, 2021. https://doi.org/10.48550/arXiv.2111.03165.","ama":"Lechner M, Žikelić Ð, Chatterjee K, Henzinger TA. Infinite time horizon safety of Bayesian neural networks. In: 35th Conference on Neural Information Processing Systems. ; 2021. doi:10.48550/arXiv.2111.03165","ieee":"M. Lechner, Ð. Žikelić, K. Chatterjee, and T. A. Henzinger, “Infinite time horizon safety of Bayesian neural networks,” in 35th Conference on Neural Information Processing Systems, Virtual, 2021.","ista":"Lechner M, Žikelić Ð, Chatterjee K, Henzinger TA. 2021. Infinite time horizon safety of Bayesian neural networks. 35th Conference on Neural Information Processing Systems. NeurIPS: Neural Information Processing Systems, Advances in Neural Information Processing Systems, .","apa":"Lechner, M., Žikelić, Ð., Chatterjee, K., & Henzinger, T. A. (2021). Infinite time horizon safety of Bayesian neural networks. In 35th Conference on Neural Information Processing Systems. Virtual. https://doi.org/10.48550/arXiv.2111.03165","mla":"Lechner, Mathias, et al. “Infinite Time Horizon Safety of Bayesian Neural Networks.” 35th Conference on Neural Information Processing Systems, 2021, doi:10.48550/arXiv.2111.03165."},"month":"12","author":[{"id":"3DC22916-F248-11E8-B48F-1D18A9856A87","first_name":"Mathias","last_name":"Lechner","full_name":"Lechner, Mathias"},{"full_name":"Žikelić, Ðorđe","last_name":"Žikelić","first_name":"Ðorđe"},{"full_name":"Chatterjee, Krishnendu","orcid":"0000-0002-4561-241X","last_name":"Chatterjee","first_name":"Krishnendu","id":"2E5DCA20-F248-11E8-B48F-1D18A9856A87"},{"orcid":"0000-0002-2985-7724","id":"40876CD8-F248-11E8-B48F-1D18A9856A87","last_name":"Henzinger","first_name":"Thomas A","full_name":"Henzinger, Thomas A"}]}