[{"publisher":"Springer Nature","doi":"10.1007/978-3-031-45329-8_17","article_processing_charge":"No","alternative_title":["LNCS"],"type":"conference","date_updated":"2025-07-14T09:09:59Z","_id":"14559","page":"357-379","quality_controlled":"1","year":"2023","project":[{"_id":"62781420-2b32-11ec-9570-8d9b63373d4d","name":"Vigilant Algorithmic Monitoring of Software","grant_number":"101020093","call_identifier":"H2020"},{"grant_number":"863818","name":"Formal Methods for Stochastic Models: Algorithms and Applications","call_identifier":"H2020","_id":"0599E47C-7A3F-11EA-A408-12923DDC885E"},{"_id":"2564DBCA-B435-11E9-9278-68D0E5697425","grant_number":"665385","name":"International IST Doctoral Program","call_identifier":"H2020"}],"status":"public","publication":"21st International Symposium on Automated Technology for Verification and Analysis","conference":{"location":"Singapore, Singapore","name":"ATVA: Automated Technology for Verification and Analysis","start_date":"2023-10-24","end_date":"2023-10-27"},"acknowledgement":"This work was supported in part by the ERC-2020-AdG 101020093, 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_published":"2023-10-22T00:00:00Z","ec_funded":1,"title":"Learning provably stabilizing neural controllers for discrete-time stochastic systems","oa_version":"None","author":[{"full_name":"Ansaripour, Matin","last_name":"Ansaripour","first_name":"Matin"},{"id":"2E5DCA20-F248-11E8-B48F-1D18A9856A87","full_name":"Chatterjee, Krishnendu","last_name":"Chatterjee","orcid":"0000-0002-4561-241X","first_name":"Krishnendu"},{"orcid":"0000-0002-2985-7724","first_name":"Thomas A","id":"40876CD8-F248-11E8-B48F-1D18A9856A87","full_name":"Henzinger, Thomas A","last_name":"Henzinger"},{"last_name":"Lechner","full_name":"Lechner, Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","first_name":"Mathias"},{"first_name":"Dorde","orcid":"0000-0002-4681-1699","id":"294AA7A6-F248-11E8-B48F-1D18A9856A87","full_name":"Zikelic, Dorde","last_name":"Zikelic"}],"scopus_import":"1","day":"22","date_created":"2023-11-19T23:00:56Z","volume":14215,"abstract":[{"text":"We consider the problem of learning control policies in discrete-time stochastic systems which guarantee that the system stabilizes within some specified stabilization region with probability 1. Our approach is based on the novel notion of stabilizing ranking supermartingales (sRSMs) that we introduce in this work. Our sRSMs overcome the limitation of methods proposed in previous works whose applicability is restricted to systems in which the stabilizing region cannot be left once entered under any control policy. We present a learning procedure that learns a control policy together with an sRSM that formally certifies probability 1 stability, both learned as neural networks. We show that this procedure can also be adapted to formally verifying that, under a given Lipschitz continuous control policy, the stochastic system stabilizes within some stabilizing region with probability 1. Our experimental evaluation shows that our learning procedure can successfully learn provably stabilizing policies in practice.","lang":"eng"}],"intvolume":"     14215","publication_identifier":{"eissn":["1611-3349"],"issn":["0302-9743"],"isbn":["9783031453281"]},"publication_status":"published","month":"10","department":[{"_id":"ToHe"},{"_id":"KrCh"}],"language":[{"iso":"eng"}],"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","citation":{"ista":"Ansaripour M, Chatterjee K, Henzinger TA, Lechner M, Zikelic D. 2023. Learning provably stabilizing neural controllers for discrete-time stochastic systems. 21st International Symposium on Automated Technology for Verification and Analysis. ATVA: Automated Technology for Verification and Analysis, LNCS, vol. 14215, 357–379.","chicago":"Ansaripour, Matin, Krishnendu Chatterjee, Thomas A Henzinger, Mathias Lechner, and Dorde Zikelic. “Learning Provably Stabilizing Neural Controllers for Discrete-Time Stochastic Systems.” In <i>21st International Symposium on Automated Technology for Verification and Analysis</i>, 14215:357–79. Springer Nature, 2023. <a href=\"https://doi.org/10.1007/978-3-031-45329-8_17\">https://doi.org/10.1007/978-3-031-45329-8_17</a>.","apa":"Ansaripour, M., Chatterjee, K., Henzinger, T. A., Lechner, M., &#38; Zikelic, D. (2023). Learning provably stabilizing neural controllers for discrete-time stochastic systems. In <i>21st International Symposium on Automated Technology for Verification and Analysis</i> (Vol. 14215, pp. 357–379). Singapore, Singapore: Springer Nature. <a href=\"https://doi.org/10.1007/978-3-031-45329-8_17\">https://doi.org/10.1007/978-3-031-45329-8_17</a>","mla":"Ansaripour, Matin, et al. “Learning Provably Stabilizing Neural Controllers for Discrete-Time Stochastic Systems.” <i>21st International Symposium on Automated Technology for Verification and Analysis</i>, vol. 14215, Springer Nature, 2023, pp. 357–79, doi:<a href=\"https://doi.org/10.1007/978-3-031-45329-8_17\">10.1007/978-3-031-45329-8_17</a>.","ama":"Ansaripour M, Chatterjee K, Henzinger TA, Lechner M, Zikelic D. Learning provably stabilizing neural controllers for discrete-time stochastic systems. In: <i>21st International Symposium on Automated Technology for Verification and Analysis</i>. Vol 14215. Springer Nature; 2023:357-379. doi:<a href=\"https://doi.org/10.1007/978-3-031-45329-8_17\">10.1007/978-3-031-45329-8_17</a>","ieee":"M. Ansaripour, K. Chatterjee, T. A. Henzinger, M. Lechner, and D. Zikelic, “Learning provably stabilizing neural controllers for discrete-time stochastic systems,” in <i>21st International Symposium on Automated Technology for Verification and Analysis</i>, Singapore, Singapore, 2023, vol. 14215, pp. 357–379.","short":"M. Ansaripour, K. Chatterjee, T.A. Henzinger, M. Lechner, D. Zikelic, in:, 21st International Symposium on Automated Technology for Verification and Analysis, Springer Nature, 2023, pp. 357–379."}},{"publisher":"Association for the Advancement of Artificial Intelligence","article_processing_charge":"No","doi":"10.1609/aaai.v37i10.26407","type":"conference","_id":"14830","date_updated":"2025-07-14T09:10:02Z","page":"11926-11935","quality_controlled":"1","related_material":{"record":[{"relation":"earlier_version","status":"public","id":"14600"}]},"external_id":{"arxiv":["2210.05308"]},"year":"2023","keyword":["General Medicine"],"publication":"Proceedings of the 37th AAAI Conference on Artificial Intelligence","status":"public","project":[{"_id":"62781420-2b32-11ec-9570-8d9b63373d4d","call_identifier":"H2020","grant_number":"101020093","name":"Vigilant Algorithmic Monitoring of Software"},{"name":"Formal Methods for Stochastic Models: Algorithms and Applications","grant_number":"863818","call_identifier":"H2020","_id":"0599E47C-7A3F-11EA-A408-12923DDC885E"},{"call_identifier":"H2020","name":"International IST Doctoral Program","grant_number":"665385","_id":"2564DBCA-B435-11E9-9278-68D0E5697425"}],"acknowledgement":"This work was supported in part by the ERC-2020-AdG 101020093, 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_published":"2023-06-26T00:00:00Z","conference":{"end_date":"2023-02-14","start_date":"2023-02-07","name":"AAAI: Conference on Artificial Intelligence","location":"Washington, DC, United States"},"ec_funded":1,"title":"Learning control policies for stochastic systems with reach-avoid guarantees","oa_version":"Preprint","day":"26","author":[{"orcid":"0000-0002-4681-1699","first_name":"Dorde","id":"294AA7A6-F248-11E8-B48F-1D18A9856A87","full_name":"Zikelic, Dorde","last_name":"Zikelic"},{"last_name":"Lechner","full_name":"Lechner, Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","first_name":"Mathias"},{"first_name":"Thomas A","orcid":"0000-0002-2985-7724","full_name":"Henzinger, Thomas A","id":"40876CD8-F248-11E8-B48F-1D18A9856A87","last_name":"Henzinger"},{"full_name":"Chatterjee, Krishnendu","id":"2E5DCA20-F248-11E8-B48F-1D18A9856A87","last_name":"Chatterjee","first_name":"Krishnendu","orcid":"0000-0002-4561-241X"}],"date_created":"2024-01-18T07:44:31Z","volume":37,"abstract":[{"lang":"eng","text":"We study the problem of learning controllers for discrete-time non-linear stochastic dynamical systems with formal reach-avoid guarantees. This work presents the first method for providing formal reach-avoid guarantees, which combine and generalize stability and safety guarantees, with a tolerable probability threshold p in [0,1] over the infinite time horizon. Our method leverages advances in machine learning literature and it represents formal certificates as neural networks. In particular, we learn a certificate in the form of a reach-avoid supermartingale (RASM), a novel notion that we introduce in this work. Our RASMs provide reachability and avoidance guarantees by imposing constraints on what can be viewed as a stochastic extension of level sets of Lyapunov functions for deterministic systems. Our approach solves several important problems -- it can be used to learn a control policy from scratch, to verify a reach-avoid specification for a fixed control policy, or to fine-tune a pre-trained policy if it does not satisfy the reach-avoid specification. We validate our approach on 3 stochastic non-linear reinforcement learning tasks."}],"intvolume":"        37","publication_identifier":{"eissn":["2374-3468"],"issn":["2159-5399"]},"publication_status":"published","arxiv":1,"month":"06","department":[{"_id":"ToHe"},{"_id":"KrCh"}],"language":[{"iso":"eng"}],"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","citation":{"ieee":"D. Zikelic, M. Lechner, T. A. Henzinger, and K. Chatterjee, “Learning control policies for stochastic systems with reach-avoid guarantees,” in <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i>, Washington, DC, United States, 2023, vol. 37, no. 10, pp. 11926–11935.","short":"D. Zikelic, M. Lechner, T.A. Henzinger, K. Chatterjee, in:, Proceedings of the 37th AAAI Conference on Artificial Intelligence, Association for the Advancement of Artificial Intelligence, 2023, pp. 11926–11935.","ama":"Zikelic D, Lechner M, Henzinger TA, Chatterjee K. Learning control policies for stochastic systems with reach-avoid guarantees. In: <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i>. Vol 37. Association for the Advancement of Artificial Intelligence; 2023:11926-11935. doi:<a href=\"https://doi.org/10.1609/aaai.v37i10.26407\">10.1609/aaai.v37i10.26407</a>","apa":"Zikelic, D., Lechner, M., Henzinger, T. A., &#38; Chatterjee, K. (2023). Learning control policies for stochastic systems with reach-avoid guarantees. In <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i> (Vol. 37, pp. 11926–11935). Washington, DC, United States: Association for the Advancement of Artificial Intelligence. <a href=\"https://doi.org/10.1609/aaai.v37i10.26407\">https://doi.org/10.1609/aaai.v37i10.26407</a>","mla":"Zikelic, Dorde, et al. “Learning Control Policies for Stochastic Systems with Reach-Avoid Guarantees.” <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i>, vol. 37, no. 10, Association for the Advancement of Artificial Intelligence, 2023, pp. 11926–35, doi:<a href=\"https://doi.org/10.1609/aaai.v37i10.26407\">10.1609/aaai.v37i10.26407</a>.","ista":"Zikelic D, Lechner M, Henzinger TA, Chatterjee K. 2023. Learning control policies for stochastic systems with reach-avoid guarantees. Proceedings of the 37th AAAI Conference on Artificial Intelligence. AAAI: Conference on Artificial Intelligence vol. 37, 11926–11935.","chicago":"Zikelic, Dorde, Mathias Lechner, Thomas A Henzinger, and Krishnendu Chatterjee. “Learning Control Policies for Stochastic Systems with Reach-Avoid Guarantees.” In <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i>, 37:11926–35. Association for the Advancement of Artificial Intelligence, 2023. <a href=\"https://doi.org/10.1609/aaai.v37i10.26407\">https://doi.org/10.1609/aaai.v37i10.26407</a>."},"issue":"10"},{"author":[{"id":"294AA7A6-F248-11E8-B48F-1D18A9856A87","full_name":"Zikelic, Dorde","last_name":"Zikelic","orcid":"0000-0002-4681-1699","first_name":"Dorde"},{"last_name":"Lechner","full_name":"Lechner, Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","first_name":"Mathias"},{"full_name":"Verma, Abhinav","id":"a235593c-d7fa-11eb-a0c5-b22ca3c66ee6","last_name":"Verma","first_name":"Abhinav"},{"orcid":"0000-0002-4561-241X","first_name":"Krishnendu","last_name":"Chatterjee","full_name":"Chatterjee, Krishnendu","id":"2E5DCA20-F248-11E8-B48F-1D18A9856A87"},{"full_name":"Henzinger, Thomas A","id":"40876CD8-F248-11E8-B48F-1D18A9856A87","last_name":"Henzinger","first_name":"Thomas A","orcid":"0000-0002-2985-7724"}],"article_processing_charge":"No","day":"15","title":"Compositional policy learning in stochastic control systems with formal guarantees","oa_version":"Preprint","date_updated":"2025-07-14T09:10:04Z","_id":"15023","type":"conference","date_created":"2024-02-25T09:23:24Z","abstract":[{"lang":"eng","text":"Reinforcement learning has shown promising results in learning neural network policies for complicated control tasks. However, the lack of formal guarantees about the behavior of such policies remains an impediment to their deployment. We propose a novel method for learning a composition of neural network policies in stochastic environments, along with a formal certificate which guarantees that a specification over the policy's behavior is satisfied with the desired probability. Unlike prior work on verifiable RL, our approach leverages the compositional nature of logical specifications provided in SpectRL, to learn over graphs of probabilistic reach-avoid specifications. The formal guarantees are provided by learning neural network policies together with reach-avoid supermartingales (RASM) for the graph’s sub-tasks and then composing them into a global policy. We also derive a tighter lower bound compared to previous work on the probability of reach-avoidance implied by a RASM, which is required to find a compositional policy with an acceptable probabilistic threshold for complex tasks with multiple edge policies. We implement a prototype of our approach and evaluate it on a Stochastic Nine Rooms environment."}],"main_file_link":[{"url":"https://doi.org/10.48550/arXiv.2312.01456","open_access":"1"}],"quality_controlled":"1","publication_status":"epub_ahead","year":"2023","external_id":{"arxiv":["2312.01456"]},"month":"12","arxiv":1,"department":[{"_id":"ToHe"},{"_id":"KrCh"}],"oa":1,"project":[{"_id":"0599E47C-7A3F-11EA-A408-12923DDC885E","call_identifier":"H2020","grant_number":"863818","name":"Formal Methods for Stochastic Models: Algorithms and Applications"},{"call_identifier":"H2020","name":"Vigilant Algorithmic Monitoring of Software","grant_number":"101020093","_id":"62781420-2b32-11ec-9570-8d9b63373d4d"}],"language":[{"iso":"eng"}],"status":"public","publication":"37th Conference on Neural Information Processing Systems","ec_funded":1,"citation":{"apa":"Zikelic, D., Lechner, M., Verma, A., Chatterjee, K., &#38; Henzinger, T. A. (2023). Compositional policy learning in stochastic control systems with formal guarantees. In <i>37th Conference on Neural Information Processing Systems</i>. New Orleans, LO, United States.","mla":"Zikelic, Dorde, et al. “Compositional Policy Learning in Stochastic Control Systems with Formal Guarantees.” <i>37th Conference on Neural Information Processing Systems</i>, 2023.","chicago":"Zikelic, Dorde, Mathias Lechner, Abhinav Verma, Krishnendu Chatterjee, and Thomas A Henzinger. “Compositional Policy Learning in Stochastic Control Systems with Formal Guarantees.” In <i>37th Conference on Neural Information Processing Systems</i>, 2023.","ista":"Zikelic D, Lechner M, Verma A, Chatterjee K, Henzinger TA. 2023. Compositional policy learning in stochastic control systems with formal guarantees. 37th Conference on Neural Information Processing Systems. NeurIPS: Neural Information Processing Systems.","ieee":"D. Zikelic, M. Lechner, A. Verma, K. Chatterjee, and T. A. Henzinger, “Compositional policy learning in stochastic control systems with formal guarantees,” in <i>37th Conference on Neural Information Processing Systems</i>, New Orleans, LO, United States, 2023.","short":"D. Zikelic, M. Lechner, A. Verma, K. Chatterjee, T.A. Henzinger, in:, 37th Conference on Neural Information Processing Systems, 2023.","ama":"Zikelic D, Lechner M, Verma A, Chatterjee K, Henzinger TA. Compositional policy learning in stochastic control systems with formal guarantees. In: <i>37th Conference on Neural Information Processing Systems</i>. ; 2023."},"date_published":"2023-12-15T00:00:00Z","conference":{"location":"New Orleans, LO, United States","start_date":"2023-12-10","end_date":"2023-12-16","name":"NeurIPS: Neural Information Processing Systems"},"acknowledgement":"This work was supported in part by the ERC-2020-AdG 101020093 (VAMOS) and the ERC-2020-\r\nCoG 863818 (FoRM-SMArt).","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87"},{"month":"04","department":[{"_id":"KrCh"},{"_id":"ToHe"}],"file":[{"date_updated":"2023-06-19T08:29:30Z","creator":"dernst","file_size":528455,"date_created":"2023-06-19T08:29:30Z","file_id":"13150","access_level":"open_access","content_type":"application/pdf","success":1,"file_name":"2023_LNCS_Chatterjee.pdf","checksum":"3d8a8bb24d211bc83360dfc2fd744307","relation":"main_file"}],"oa":1,"language":[{"iso":"eng"}],"citation":{"ista":"Chatterjee K, Henzinger TA, Lechner M, Zikelic D. 2023. A learner-verifier framework for neural network controllers and certificates of stochastic systems. Tools and Algorithms for the Construction and Analysis of Systems . TACAS: Tools and Algorithms for the Construction and Analysis of Systems, LNCS, vol. 13993, 3–25.","chicago":"Chatterjee, Krishnendu, Thomas A Henzinger, Mathias Lechner, and Dorde Zikelic. “A Learner-Verifier Framework for Neural Network Controllers and Certificates of Stochastic Systems.” In <i>Tools and Algorithms for the Construction and Analysis of Systems </i>, 13993:3–25. Springer Nature, 2023. <a href=\"https://doi.org/10.1007/978-3-031-30823-9_1\">https://doi.org/10.1007/978-3-031-30823-9_1</a>.","mla":"Chatterjee, Krishnendu, et al. “A Learner-Verifier Framework for Neural Network Controllers and Certificates of Stochastic Systems.” <i>Tools and Algorithms for the Construction and Analysis of Systems </i>, vol. 13993, Springer Nature, 2023, pp. 3–25, doi:<a href=\"https://doi.org/10.1007/978-3-031-30823-9_1\">10.1007/978-3-031-30823-9_1</a>.","apa":"Chatterjee, K., Henzinger, T. A., Lechner, M., &#38; Zikelic, D. (2023). A learner-verifier framework for neural network controllers and certificates of stochastic systems. In <i>Tools and Algorithms for the Construction and Analysis of Systems </i> (Vol. 13993, pp. 3–25). Paris, France: Springer Nature. <a href=\"https://doi.org/10.1007/978-3-031-30823-9_1\">https://doi.org/10.1007/978-3-031-30823-9_1</a>","ama":"Chatterjee K, Henzinger TA, Lechner M, Zikelic D. A learner-verifier framework for neural network controllers and certificates of stochastic systems. In: <i>Tools and Algorithms for the Construction and Analysis of Systems </i>. Vol 13993. Springer Nature; 2023:3-25. doi:<a href=\"https://doi.org/10.1007/978-3-031-30823-9_1\">10.1007/978-3-031-30823-9_1</a>","short":"K. Chatterjee, T.A. Henzinger, M. Lechner, D. Zikelic, in:, Tools and Algorithms for the Construction and Analysis of Systems , Springer Nature, 2023, pp. 3–25.","ieee":"K. Chatterjee, T. A. Henzinger, M. Lechner, and D. Zikelic, “A learner-verifier framework for neural network controllers and certificates of stochastic systems,” in <i>Tools and Algorithms for the Construction and Analysis of Systems </i>, Paris, France, 2023, vol. 13993, pp. 3–25."},"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","author":[{"full_name":"Chatterjee, Krishnendu","id":"2E5DCA20-F248-11E8-B48F-1D18A9856A87","last_name":"Chatterjee","orcid":"0000-0002-4561-241X","first_name":"Krishnendu"},{"orcid":"0000-0002-2985-7724","first_name":"Thomas A","id":"40876CD8-F248-11E8-B48F-1D18A9856A87","full_name":"Henzinger, Thomas A","last_name":"Henzinger"},{"last_name":"Lechner","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","full_name":"Lechner, Mathias","first_name":"Mathias"},{"first_name":"Dorde","orcid":"0000-0002-4681-1699","id":"294AA7A6-F248-11E8-B48F-1D18A9856A87","full_name":"Zikelic, Dorde","last_name":"Zikelic"}],"day":"22","scopus_import":"1","title":"A learner-verifier framework for neural network controllers and certificates of stochastic systems","oa_version":"Published Version","volume":13993,"date_created":"2023-06-18T22:00:47Z","has_accepted_license":"1","tmp":{"name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode","image":"/images/cc_by.png","short":"CC BY (4.0)"},"abstract":[{"lang":"eng","text":"Reinforcement learning has received much attention for learning controllers of deterministic systems. We consider a learner-verifier framework for stochastic control systems and survey recent methods that formally guarantee a conjunction of reachability and safety properties. Given a property and a lower bound on the probability of the property being satisfied, our framework jointly learns a control policy and a formal certificate to ensure the satisfaction of the property with a desired probability threshold. Both the control policy and the formal certificate are continuous functions from states to reals, which are learned as parameterized neural networks. While in the deterministic case, the certificates are invariant and barrier functions for safety, or Lyapunov and ranking functions for liveness, in the stochastic case the certificates are supermartingales. For certificate verification, we use interval arithmetic abstract interpretation to bound the expected values of neural network functions."}],"intvolume":"     13993","publication_status":"published","publication_identifier":{"issn":["0302-9743"],"isbn":["9783031308222"],"eissn":["1611-3349"]},"file_date_updated":"2023-06-19T08:29:30Z","year":"2023","project":[{"_id":"0599E47C-7A3F-11EA-A408-12923DDC885E","name":"Formal Methods for Stochastic Models: Algorithms and Applications","grant_number":"863818","call_identifier":"H2020"},{"_id":"2564DBCA-B435-11E9-9278-68D0E5697425","call_identifier":"H2020","name":"International IST Doctoral Program","grant_number":"665385"}],"publication":"Tools and Algorithms for the Construction and Analysis of Systems ","status":"public","ec_funded":1,"date_published":"2023-04-22T00:00:00Z","acknowledgement":"This work was supported in part by the ERC-2020-AdG 101020093, 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.","conference":{"name":"TACAS: Tools and Algorithms for the Construction and Analysis of Systems","start_date":"2023-04-22","end_date":"2023-04-27","location":"Paris, France"},"doi":"10.1007/978-3-031-30823-9_1","alternative_title":["LNCS"],"article_processing_charge":"No","publisher":"Springer Nature","date_updated":"2025-07-14T09:09:52Z","_id":"13142","type":"conference","page":"3-25","ddc":["000"],"quality_controlled":"1"},{"author":[{"first_name":"Mathias","full_name":"Lechner, Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","last_name":"Lechner"},{"full_name":"Zikelic, Dorde","id":"294AA7A6-F248-11E8-B48F-1D18A9856A87","last_name":"Zikelic","orcid":"0000-0002-4681-1699","first_name":"Dorde"},{"last_name":"Chatterjee","full_name":"Chatterjee, Krishnendu","id":"2E5DCA20-F248-11E8-B48F-1D18A9856A87","orcid":"0000-0002-4561-241X","first_name":"Krishnendu"},{"first_name":"Thomas A","orcid":"0000-0002-2985-7724","last_name":"Henzinger","id":"40876CD8-F248-11E8-B48F-1D18A9856A87","full_name":"Henzinger, Thomas A"},{"first_name":"Daniela","last_name":"Rus","full_name":"Rus, Daniela"}],"day":"26","scopus_import":"1","oa_version":"Preprint","title":"Quantization-aware interval bound propagation for training certifiably robust quantized neural networks","volume":37,"date_created":"2023-08-27T22:01:17Z","intvolume":"        37","abstract":[{"text":"We study the problem of training and certifying adversarially robust quantized neural networks (QNNs). Quantization is a technique for making neural networks more efficient by running them using low-bit integer arithmetic and is therefore commonly adopted in industry. Recent work has shown that floating-point neural networks that have been verified to be robust can become vulnerable to adversarial attacks after quantization, and certification of the quantized representation is necessary to guarantee robustness. In this work, we present quantization-aware interval bound propagation (QA-IBP), a novel method for training robust QNNs. Inspired by advances in robust learning of non-quantized networks, our training algorithm computes the gradient of an abstract representation of the actual network. Unlike existing approaches, our method can handle the discrete semantics of QNNs. Based on QA-IBP, we also develop a complete verification procedure for verifying the adversarial robustness of QNNs, which is guaranteed to terminate and produce a correct answer. Compared to existing approaches, the key advantage of our verification procedure is that it runs entirely on GPU or other accelerator devices. We demonstrate experimentally that our approach significantly outperforms existing methods and establish the new state-of-the-art for training and certifying the robustness of QNNs.","lang":"eng"}],"publication_identifier":{"isbn":["9781577358800"]},"publication_status":"published","month":"06","arxiv":1,"department":[{"_id":"ToHe"},{"_id":"KrCh"}],"oa":1,"language":[{"iso":"eng"}],"issue":"12","citation":{"chicago":"Lechner, Mathias, Dorde Zikelic, Krishnendu Chatterjee, Thomas A Henzinger, and Daniela Rus. “Quantization-Aware Interval Bound Propagation for Training Certifiably Robust Quantized Neural Networks.” In <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i>, 37:14964–73. Association for the Advancement of Artificial Intelligence, 2023. <a href=\"https://doi.org/10.1609/aaai.v37i12.26747\">https://doi.org/10.1609/aaai.v37i12.26747</a>.","ista":"Lechner M, Zikelic D, Chatterjee K, Henzinger TA, Rus D. 2023. Quantization-aware interval bound propagation for training certifiably robust quantized neural networks. Proceedings of the 37th AAAI Conference on Artificial Intelligence. AAAI: Conference on Artificial Intelligence vol. 37, 14964–14973.","apa":"Lechner, M., Zikelic, D., Chatterjee, K., Henzinger, T. A., &#38; Rus, D. (2023). Quantization-aware interval bound propagation for training certifiably robust quantized neural networks. In <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i> (Vol. 37, pp. 14964–14973). Washington, DC, United States: Association for the Advancement of Artificial Intelligence. <a href=\"https://doi.org/10.1609/aaai.v37i12.26747\">https://doi.org/10.1609/aaai.v37i12.26747</a>","mla":"Lechner, Mathias, et al. “Quantization-Aware Interval Bound Propagation for Training Certifiably Robust Quantized Neural Networks.” <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i>, vol. 37, no. 12, Association for the Advancement of Artificial Intelligence, 2023, pp. 14964–73, doi:<a href=\"https://doi.org/10.1609/aaai.v37i12.26747\">10.1609/aaai.v37i12.26747</a>.","ama":"Lechner M, Zikelic D, Chatterjee K, Henzinger TA, Rus D. Quantization-aware interval bound propagation for training certifiably robust quantized neural networks. In: <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i>. Vol 37. Association for the Advancement of Artificial Intelligence; 2023:14964-14973. doi:<a href=\"https://doi.org/10.1609/aaai.v37i12.26747\">10.1609/aaai.v37i12.26747</a>","ieee":"M. Lechner, D. Zikelic, K. Chatterjee, T. A. Henzinger, and D. Rus, “Quantization-aware interval bound propagation for training certifiably robust quantized neural networks,” in <i>Proceedings of the 37th AAAI Conference on Artificial Intelligence</i>, Washington, DC, United States, 2023, vol. 37, no. 12, pp. 14964–14973.","short":"M. Lechner, D. Zikelic, K. Chatterjee, T.A. Henzinger, D. Rus, in:, Proceedings of the 37th AAAI Conference on Artificial Intelligence, Association for the Advancement of Artificial Intelligence, 2023, pp. 14964–14973."},"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","doi":"10.1609/aaai.v37i12.26747","article_processing_charge":"No","publisher":"Association for the Advancement of Artificial Intelligence","date_updated":"2025-07-14T09:09:56Z","_id":"14242","type":"conference","page":"14964-14973","main_file_link":[{"open_access":"1","url":"https://doi.org/10.48550/arXiv.2211.16187"}],"quality_controlled":"1","year":"2023","external_id":{"arxiv":["2211.16187"]},"project":[{"_id":"62781420-2b32-11ec-9570-8d9b63373d4d","call_identifier":"H2020","name":"Vigilant Algorithmic Monitoring of Software","grant_number":"101020093"},{"call_identifier":"H2020","grant_number":"863818","name":"Formal Methods for Stochastic Models: Algorithms and Applications","_id":"0599E47C-7A3F-11EA-A408-12923DDC885E"},{"_id":"2564DBCA-B435-11E9-9278-68D0E5697425","grant_number":"665385","name":"International IST Doctoral Program","call_identifier":"H2020"}],"publication":"Proceedings of the 37th AAAI Conference on Artificial Intelligence","status":"public","ec_funded":1,"date_published":"2023-06-26T00:00:00Z","acknowledgement":"This work was supported in part by the ERC-2020-AdG 101020093, 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. Research was sponsored by the United\r\nStates Air Force Research Laboratory and the United States Air Force Artificial Intelligence Accelerator and was accomplished under Cooperative Agreement Number FA8750-19-2-\r\n1000. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied,\r\nof the United States Air Force or the U.S. Government. The U.S. Government is authorized to reproduce and distribute reprints for Government purposes notwithstanding any copyright\r\nnotation herein. The research was also funded in part by the AI2050 program at Schmidt Futures (Grant G-22-63172) and Capgemini SE.","conference":{"end_date":"2023-02-14","start_date":"2023-02-07","name":"AAAI: Conference on Artificial Intelligence","location":"Washington, DC, United States"}},{"author":[{"full_name":"Lechner, Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","last_name":"Lechner","first_name":"Mathias"},{"first_name":"Alexander","full_name":"Amini, Alexander","last_name":"Amini"},{"first_name":"Daniela","last_name":"Rus","full_name":"Rus, Daniela"},{"id":"40876CD8-F248-11E8-B48F-1D18A9856A87","full_name":"Henzinger, Thomas A","last_name":"Henzinger","first_name":"Thomas A","orcid":"0000-0002-2985-7724"}],"scopus_import":"1","day":"01","oa_version":"Published Version","title":"Revisiting the adversarial robustness-accuracy tradeoff in robot learning","volume":8,"article_type":"original","date_created":"2023-03-05T23:01:04Z","has_accepted_license":"1","intvolume":"         8","tmp":{"name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode","image":"/images/cc_by.png","short":"CC BY (4.0)"},"abstract":[{"text":"Adversarial training (i.e., training on adversarially perturbed input data) is a well-studied method for making neural networks robust to potential adversarial attacks during inference. However, the improved robustness does not come for free but rather is accompanied by a decrease in overall model accuracy and performance. Recent work has shown that, in practical robot learning applications, the effects of adversarial training do not pose a fair trade-off but inflict a net loss when measured in holistic robot performance. This work revisits the robustness-accuracy trade-off in robot learning by systematically analyzing if recent advances in robust training methods and theory in conjunction with adversarial robot learning, are capable of making adversarial training suitable for real-world robot applications. We evaluate three different robot learning tasks ranging from autonomous driving in a high-fidelity environment amenable to sim-to-real deployment to mobile robot navigation and gesture recognition. Our results demonstrate that, while these techniques make incremental improvements on the trade-off on a relative scale, the negative impact on the nominal accuracy caused by adversarial training still outweighs the improved robustness by an order of magnitude. We conclude that although progress is happening, further advances in robust learning methods are necessary before they can benefit robot learning tasks in practice.","lang":"eng"}],"publication_status":"published","publication_identifier":{"eissn":["2377-3766"]},"file_date_updated":"2023-03-07T12:22:23Z","month":"03","arxiv":1,"department":[{"_id":"ToHe"}],"file":[{"date_updated":"2023-03-07T12:22:23Z","creator":"cchlebak","date_created":"2023-03-07T12:22:23Z","file_size":944052,"file_id":"12714","content_type":"application/pdf","access_level":"open_access","file_name":"2023_IEEERobAutLetters_Lechner.pdf","success":1,"checksum":"5a75dcd326ea66685de2b1aaec259e85","relation":"main_file"}],"oa":1,"language":[{"iso":"eng"}],"issue":"3","citation":{"ama":"Lechner M, Amini A, Rus D, Henzinger TA. Revisiting the adversarial robustness-accuracy tradeoff in robot learning. <i>IEEE Robotics and Automation Letters</i>. 2023;8(3):1595-1602. doi:<a href=\"https://doi.org/10.1109/LRA.2023.3240930\">10.1109/LRA.2023.3240930</a>","ieee":"M. Lechner, A. Amini, D. Rus, and T. A. Henzinger, “Revisiting the adversarial robustness-accuracy tradeoff in robot learning,” <i>IEEE Robotics and Automation Letters</i>, vol. 8, no. 3. Institute of Electrical and Electronics Engineers, pp. 1595–1602, 2023.","short":"M. Lechner, A. Amini, D. Rus, T.A. Henzinger, IEEE Robotics and Automation Letters 8 (2023) 1595–1602.","ista":"Lechner M, Amini A, Rus D, Henzinger TA. 2023. Revisiting the adversarial robustness-accuracy tradeoff in robot learning. IEEE Robotics and Automation Letters. 8(3), 1595–1602.","chicago":"Lechner, Mathias, Alexander Amini, Daniela Rus, and Thomas A Henzinger. “Revisiting the Adversarial Robustness-Accuracy Tradeoff in Robot Learning.” <i>IEEE Robotics and Automation Letters</i>. Institute of Electrical and Electronics Engineers, 2023. <a href=\"https://doi.org/10.1109/LRA.2023.3240930\">https://doi.org/10.1109/LRA.2023.3240930</a>.","apa":"Lechner, M., Amini, A., Rus, D., &#38; Henzinger, T. A. (2023). Revisiting the adversarial robustness-accuracy tradeoff in robot learning. <i>IEEE Robotics and Automation Letters</i>. Institute of Electrical and Electronics Engineers. <a href=\"https://doi.org/10.1109/LRA.2023.3240930\">https://doi.org/10.1109/LRA.2023.3240930</a>","mla":"Lechner, Mathias, et al. “Revisiting the Adversarial Robustness-Accuracy Tradeoff in Robot Learning.” <i>IEEE Robotics and Automation Letters</i>, vol. 8, no. 3, Institute of Electrical and Electronics Engineers, 2023, pp. 1595–602, doi:<a href=\"https://doi.org/10.1109/LRA.2023.3240930\">10.1109/LRA.2023.3240930</a>."},"user_id":"4359f0d1-fa6c-11eb-b949-802e58b17ae8","doi":"10.1109/LRA.2023.3240930","article_processing_charge":"No","publisher":"Institute of Electrical and Electronics Engineers","date_updated":"2023-08-01T13:36:50Z","_id":"12704","type":"journal_article","page":"1595-1602","ddc":["000"],"quality_controlled":"1","isi":1,"year":"2023","external_id":{"isi":["000936534100012"],"arxiv":["2204.07373"]},"related_material":{"record":[{"id":"11366","relation":"earlier_version","status":"public"}]},"status":"public","publication":"IEEE Robotics and Automation Letters","acknowledgement":"We thank Christoph Lampert for inspiring this work. The\r\nviews and conclusions contained in this document are those of\r\nthe authors and should not be interpreted as representing the\r\nofficial policies, either expressed or implied, of the United States\r\nAir Force or the U.S. Government. The U.S. Government is\r\nauthorized to reproduce and distribute reprints for Government\r\npurposes notwithstanding any copyright notation herein.","date_published":"2023-03-01T00:00:00Z"},{"page":"124","ddc":["004"],"_id":"11362","date_updated":"2025-07-14T09:10:11Z","type":"dissertation","alternative_title":["ISTA Thesis"],"article_processing_charge":"No","doi":"10.15479/at:ista:11362","publisher":"Institute of Science and Technology Austria","ec_funded":1,"date_published":"2022-05-12T00:00:00Z","degree_awarded":"PhD","status":"public","project":[{"_id":"25F42A32-B435-11E9-9278-68D0E5697425","grant_number":"Z211","name":"The Wittgenstein Prize","call_identifier":"FWF"},{"call_identifier":"H2020","grant_number":"101020093","name":"Vigilant Algorithmic Monitoring of Software","_id":"62781420-2b32-11ec-9570-8d9b63373d4d"}],"keyword":["neural networks","verification","machine learning"],"year":"2022","related_material":{"record":[{"id":"11366","relation":"part_of_dissertation","status":"public"},{"id":"7808","status":"public","relation":"part_of_dissertation"},{"id":"10666","status":"public","relation":"part_of_dissertation"},{"id":"10665","relation":"part_of_dissertation","status":"public"},{"id":"10667","relation":"part_of_dissertation","status":"public"}]},"file_date_updated":"2022-05-17T15:19:39Z","publication_status":"published","publication_identifier":{"isbn":["978-3-99078-017-6"]},"has_accepted_license":"1","license":"https://creativecommons.org/licenses/by-nd/4.0/","abstract":[{"lang":"eng","text":"Deep learning has enabled breakthroughs in challenging computing problems and has emerged as the standard problem-solving tool for computer vision and natural language processing tasks.\r\nOne exception to this trend is safety-critical tasks where robustness and resilience requirements contradict the black-box nature of neural networks. \r\nTo deploy deep learning methods for these tasks, it is vital to provide guarantees on neural network agents' safety and robustness criteria. \r\nThis can be achieved by developing formal verification methods to verify the safety and robustness properties of neural networks.\r\n\r\nOur goal is to design, develop and assess safety verification methods for neural networks to improve their reliability and trustworthiness in real-world applications.\r\nThis thesis establishes techniques for the verification of compressed and adversarially trained models as well as the design of novel neural networks for verifiably safe decision-making.\r\n\r\nFirst, we establish the problem of verifying quantized neural networks. Quantization is a technique that trades numerical precision for the computational efficiency of running a neural network and is widely adopted in industry.\r\nWe show that neglecting the reduced precision when verifying a neural network can lead to wrong conclusions about the robustness and safety of the network, highlighting that novel techniques for quantized network verification are necessary. We introduce several bit-exact verification methods explicitly designed for quantized neural networks and experimentally confirm on realistic networks that the network's robustness and other formal properties are affected by the quantization.\r\n\r\nFurthermore, we perform a case study providing evidence that adversarial training, a standard technique for making neural networks more robust, has detrimental effects on the network's performance. This robustness-accuracy tradeoff has been studied before regarding the accuracy obtained on classification datasets where each data point is independent of all other data points. On the other hand, we investigate the tradeoff empirically in robot learning settings where a both, a high accuracy and a high robustness, are desirable.\r\nOur results suggest that the negative side-effects of adversarial training outweigh its robustness benefits in practice.\r\n\r\nFinally, we consider the problem of verifying safety when running a Bayesian neural network policy in a feedback loop with systems over the infinite time horizon. Bayesian neural networks are probabilistic models for learning uncertainties in the data and are therefore often used on robotic and healthcare applications where data is inherently stochastic.\r\nWe introduce a method for recalibrating Bayesian neural networks so that they yield probability distributions over safe decisions only.\r\nOur method learns a safety certificate that guarantees safety over the infinite time horizon to determine which decisions are safe in every possible state of the system.\r\nWe demonstrate the effectiveness of our approach on a series of reinforcement learning benchmarks."}],"tmp":{"legal_code_url":"https://creativecommons.org/licenses/by-nd/4.0/legalcode","name":"Creative Commons Attribution-NoDerivatives 4.0 International (CC BY-ND 4.0)","image":"/image/cc_by_nd.png","short":"CC BY-ND (4.0)"},"date_created":"2022-05-12T07:14:01Z","day":"12","author":[{"last_name":"Lechner","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","full_name":"Lechner, Mathias","first_name":"Mathias"}],"oa_version":"Published Version","title":"Learning verifiable representations","citation":{"ista":"Lechner M. 2022. Learning verifiable representations. Institute of Science and Technology Austria.","chicago":"Lechner, Mathias. “Learning Verifiable Representations.” Institute of Science and Technology Austria, 2022. <a href=\"https://doi.org/10.15479/at:ista:11362\">https://doi.org/10.15479/at:ista:11362</a>.","apa":"Lechner, M. (2022). <i>Learning verifiable representations</i>. Institute of Science and Technology Austria. <a href=\"https://doi.org/10.15479/at:ista:11362\">https://doi.org/10.15479/at:ista:11362</a>","mla":"Lechner, Mathias. <i>Learning Verifiable Representations</i>. Institute of Science and Technology Austria, 2022, doi:<a href=\"https://doi.org/10.15479/at:ista:11362\">10.15479/at:ista:11362</a>.","ama":"Lechner M. Learning verifiable representations. 2022. doi:<a href=\"https://doi.org/10.15479/at:ista:11362\">10.15479/at:ista:11362</a>","ieee":"M. Lechner, “Learning verifiable representations,” Institute of Science and Technology Austria, 2022.","short":"M. Lechner, Learning Verifiable Representations, Institute of Science and Technology Austria, 2022."},"user_id":"8b945eb4-e2f2-11eb-945a-df72226e66a9","oa":1,"language":[{"iso":"eng"}],"department":[{"_id":"GradSch"},{"_id":"ToHe"}],"file":[{"access_level":"closed","content_type":"application/zip","file_name":"src.zip","checksum":"8eefa9c7c10ca7e1a2ccdd731962a645","relation":"source_file","creator":"mlechner","date_updated":"2022-05-13T12:49:00Z","date_created":"2022-05-13T12:33:26Z","file_size":13210143,"file_id":"11378"},{"date_created":"2022-05-16T08:02:28Z","file_size":2732536,"date_updated":"2022-05-17T15:19:39Z","creator":"mlechner","file_id":"11382","file_name":"thesis_main-a2.pdf","access_level":"open_access","content_type":"application/pdf","relation":"main_file","checksum":"1b9e1e5a9a83ed9d89dad2f5133dc026"}],"supervisor":[{"orcid":"0000-0002-2985-7724","first_name":"Thomas A","last_name":"Henzinger","id":"40876CD8-F248-11E8-B48F-1D18A9856A87","full_name":"Henzinger, Thomas A"}],"month":"05"},{"oa_version":"Preprint","title":"Revisiting the adversarial robustness-accuracy tradeoff in robot learning","article_processing_charge":"No","day":"15","doi":"10.48550/arXiv.2204.07373","author":[{"first_name":"Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","full_name":"Lechner, Mathias","last_name":"Lechner"},{"full_name":"Amini, Alexander","last_name":"Amini","first_name":"Alexander"},{"last_name":"Rus","full_name":"Rus, Daniela","first_name":"Daniela"},{"orcid":"0000-0002-2985-7724","first_name":"Thomas A","full_name":"Henzinger, Thomas A","id":"40876CD8-F248-11E8-B48F-1D18A9856A87","last_name":"Henzinger"}],"date_created":"2022-05-12T13:20:17Z","type":"preprint","_id":"11366","date_updated":"2023-08-01T13:36:50Z","abstract":[{"text":"Adversarial training (i.e., training on adversarially perturbed input data) is a well-studied method for making neural networks robust to potential adversarial attacks during inference. However, the improved robustness does not\r\ncome for free but rather is accompanied by a decrease in overall model accuracy and performance. Recent work has shown that, in practical robot learning applications, the effects of adversarial training do not pose a fair trade-off\r\nbut inflict a net loss when measured in holistic robot performance. This work revisits the robustness-accuracy trade-off in robot learning by systematically analyzing if recent advances in robust training methods and theory in\r\nconjunction with adversarial robot learning can make adversarial training suitable for real-world robot applications. We evaluate a wide variety of robot learning tasks ranging from autonomous driving in a high-fidelity environment\r\namenable to sim-to-real deployment, to mobile robot gesture recognition. Our results demonstrate that, while these techniques make incremental improvements on the trade-off on a relative scale, the negative side-effects caused by\r\nadversarial training still outweigh the improvements by an order of magnitude. We conclude that more substantial advances in robust learning methods are necessary before they can benefit robot learning tasks in practice.","lang":"eng"}],"publication_status":"submitted","main_file_link":[{"url":"https://doi.org/10.48550/arXiv.2204.07373","open_access":"1"}],"month":"04","related_material":{"record":[{"relation":"dissertation_contains","status":"public","id":"11362"},{"id":"12704","relation":"later_version","status":"public"}]},"arxiv":1,"external_id":{"arxiv":["2204.07373"]},"year":"2022","article_number":"2204.07373","department":[{"_id":"ToHe"}],"language":[{"iso":"eng"}],"publication":"arXiv","status":"public","project":[{"_id":"62781420-2b32-11ec-9570-8d9b63373d4d","call_identifier":"H2020","name":"Vigilant Algorithmic Monitoring of Software","grant_number":"101020093"}],"oa":1,"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","acknowledgement":"This work was supported in parts by the ERC-2020-AdG 101020093, National Science Foundation (NSF), and JP\r\nMorgan Graduate Fellowships. We thank Christoph Lampert for inspiring this work.\r\n","date_published":"2022-04-15T00:00:00Z","citation":{"ama":"Lechner M, Amini A, Rus D, Henzinger TA. Revisiting the adversarial robustness-accuracy tradeoff in robot learning. <i>arXiv</i>. doi:<a href=\"https://doi.org/10.48550/arXiv.2204.07373\">10.48550/arXiv.2204.07373</a>","ieee":"M. Lechner, A. Amini, D. Rus, and T. A. Henzinger, “Revisiting the adversarial robustness-accuracy tradeoff in robot learning,” <i>arXiv</i>. .","short":"M. Lechner, A. Amini, D. Rus, T.A. Henzinger, ArXiv (n.d.).","chicago":"Lechner, Mathias, Alexander Amini, Daniela Rus, and Thomas A Henzinger. “Revisiting the Adversarial Robustness-Accuracy Tradeoff in Robot Learning.” <i>ArXiv</i>, n.d. <a href=\"https://doi.org/10.48550/arXiv.2204.07373\">https://doi.org/10.48550/arXiv.2204.07373</a>.","ista":"Lechner M, Amini A, Rus D, Henzinger TA. Revisiting the adversarial robustness-accuracy tradeoff in robot learning. arXiv, 2204.07373.","apa":"Lechner, M., Amini, A., Rus, D., &#38; Henzinger, T. A. (n.d.). Revisiting the adversarial robustness-accuracy tradeoff in robot learning. <i>arXiv</i>. <a href=\"https://doi.org/10.48550/arXiv.2204.07373\">https://doi.org/10.48550/arXiv.2204.07373</a>","mla":"Lechner, Mathias, et al. “Revisiting the Adversarial Robustness-Accuracy Tradeoff in Robot Learning.” <i>ArXiv</i>, 2204.07373, doi:<a href=\"https://doi.org/10.48550/arXiv.2204.07373\">10.48550/arXiv.2204.07373</a>."},"ec_funded":1},{"oa":1,"project":[{"_id":"0599E47C-7A3F-11EA-A408-12923DDC885E","name":"Formal Methods for Stochastic Models: Algorithms and Applications","grant_number":"863818","call_identifier":"H2020"},{"call_identifier":"H2020","name":"Vigilant Algorithmic Monitoring of Software","grant_number":"101020093","_id":"62781420-2b32-11ec-9570-8d9b63373d4d"},{"call_identifier":"H2020","grant_number":"665385","name":"International IST Doctoral Program","_id":"2564DBCA-B435-11E9-9278-68D0E5697425"}],"language":[{"iso":"eng"}],"publication":"arXiv","status":"public","ec_funded":1,"citation":{"mla":"Zikelic, Dorde, et al. “Learning Control Policies for Stochastic Systems with Reach-Avoid Guarantees.” <i>ArXiv</i>, doi:<a href=\"https://doi.org/10.48550/ARXIV.2210.05308\">10.48550/ARXIV.2210.05308</a>.","apa":"Zikelic, D., Lechner, M., Henzinger, T. A., &#38; Chatterjee, K. (n.d.). Learning control policies for stochastic systems with reach-avoid guarantees. <i>arXiv</i>. <a href=\"https://doi.org/10.48550/ARXIV.2210.05308\">https://doi.org/10.48550/ARXIV.2210.05308</a>","chicago":"Zikelic, Dorde, Mathias Lechner, Thomas A Henzinger, and Krishnendu Chatterjee. “Learning Control Policies for Stochastic Systems with Reach-Avoid Guarantees.” <i>ArXiv</i>, n.d. <a href=\"https://doi.org/10.48550/ARXIV.2210.05308\">https://doi.org/10.48550/ARXIV.2210.05308</a>.","ista":"Zikelic D, Lechner M, Henzinger TA, Chatterjee K. Learning control policies for stochastic systems with reach-avoid guarantees. arXiv, <a href=\"https://doi.org/10.48550/ARXIV.2210.05308\">10.48550/ARXIV.2210.05308</a>.","short":"D. Zikelic, M. Lechner, T.A. Henzinger, K. Chatterjee, ArXiv (n.d.).","ieee":"D. Zikelic, M. Lechner, T. A. Henzinger, and K. Chatterjee, “Learning control policies for stochastic systems with reach-avoid guarantees,” <i>arXiv</i>. .","ama":"Zikelic D, Lechner M, Henzinger TA, Chatterjee K. Learning control policies for stochastic systems with reach-avoid guarantees. <i>arXiv</i>. doi:<a href=\"https://doi.org/10.48550/ARXIV.2210.05308\">10.48550/ARXIV.2210.05308</a>"},"date_published":"2022-11-29T00:00:00Z","user_id":"8b945eb4-e2f2-11eb-945a-df72226e66a9","year":"2022","external_id":{"arxiv":["2210.05308"]},"related_material":{"record":[{"status":"public","relation":"dissertation_contains","id":"14539"},{"relation":"later_version","status":"public","id":"14830"}]},"arxiv":1,"month":"11","department":[{"_id":"KrCh"},{"_id":"ToHe"}],"tmp":{"short":"CC BY-SA (4.0)","image":"/images/cc_by_sa.png","name":"Creative Commons Attribution-ShareAlike 4.0 International Public License (CC BY-SA 4.0)","legal_code_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode"},"license":"https://creativecommons.org/licenses/by-sa/4.0/","abstract":[{"text":"We study the problem of learning controllers for discrete-time non-linear stochastic dynamical systems with formal reach-avoid guarantees. This work presents the first method for providing formal reach-avoid guarantees, which combine and generalize stability and safety guarantees, with a tolerable probability threshold $p\\in[0,1]$ over the infinite time horizon. Our method leverages advances in machine learning literature and it represents formal certificates as neural networks. In particular, we learn a certificate in the form of a reach-avoid supermartingale (RASM), a novel notion that we introduce in this work. Our RASMs provide reachability and avoidance guarantees by imposing constraints on what can be viewed as a stochastic extension of level sets of Lyapunov functions for deterministic systems. Our approach solves several important problems -- it can be used to learn a control policy from scratch, to verify a reach-avoid specification for a fixed control policy, or to fine-tune a pre-trained policy if it does not satisfy the reach-avoid specification. We validate our approach on $3$ stochastic non-linear reinforcement learning tasks.","lang":"eng"}],"main_file_link":[{"url":"https://arxiv.org/abs/2210.05308","open_access":"1"}],"publication_status":"submitted","doi":"10.48550/ARXIV.2210.05308","author":[{"orcid":"0000-0002-4681-1699","first_name":"Dorde","last_name":"Zikelic","full_name":"Zikelic, Dorde","id":"294AA7A6-F248-11E8-B48F-1D18A9856A87"},{"full_name":"Lechner, Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","last_name":"Lechner","first_name":"Mathias"},{"last_name":"Henzinger","full_name":"Henzinger, Thomas A","id":"40876CD8-F248-11E8-B48F-1D18A9856A87","first_name":"Thomas A","orcid":"0000-0002-2985-7724"},{"first_name":"Krishnendu","orcid":"0000-0002-4561-241X","full_name":"Chatterjee, Krishnendu","id":"2E5DCA20-F248-11E8-B48F-1D18A9856A87","last_name":"Chatterjee"}],"day":"29","article_processing_charge":"No","oa_version":"Preprint","title":"Learning control policies for stochastic systems with reach-avoid guarantees","date_updated":"2025-07-14T09:10:02Z","_id":"14600","type":"preprint","date_created":"2023-11-24T13:10:09Z"},{"project":[{"_id":"62781420-2b32-11ec-9570-8d9b63373d4d","call_identifier":"H2020","name":"Vigilant Algorithmic Monitoring of Software","grant_number":"101020093"},{"grant_number":"863818","name":"Formal Methods for Stochastic Models: Algorithms and Applications","call_identifier":"H2020","_id":"0599E47C-7A3F-11EA-A408-12923DDC885E"},{"_id":"2564DBCA-B435-11E9-9278-68D0E5697425","call_identifier":"H2020","name":"International IST Doctoral Program","grant_number":"665385"}],"publication":"arXiv","status":"public","language":[{"iso":"eng"}],"oa":1,"date_published":"2022-05-24T00:00:00Z","user_id":"8b945eb4-e2f2-11eb-945a-df72226e66a9","ec_funded":1,"citation":{"ama":"Zikelic D, Lechner M, Chatterjee K, Henzinger TA. Learning stabilizing policies in stochastic control systems. <i>arXiv</i>. doi:<a href=\"https://doi.org/10.48550/arXiv.2205.11991\">10.48550/arXiv.2205.11991</a>","ieee":"D. Zikelic, M. Lechner, K. Chatterjee, and T. A. Henzinger, “Learning stabilizing policies in stochastic control systems,” <i>arXiv</i>. .","short":"D. Zikelic, M. Lechner, K. Chatterjee, T.A. Henzinger, ArXiv (n.d.).","ista":"Zikelic D, Lechner M, Chatterjee K, Henzinger TA. Learning stabilizing policies in stochastic control systems. arXiv, <a href=\"https://doi.org/10.48550/arXiv.2205.11991\">10.48550/arXiv.2205.11991</a>.","chicago":"Zikelic, Dorde, Mathias Lechner, Krishnendu Chatterjee, and Thomas A Henzinger. “Learning Stabilizing Policies in Stochastic Control Systems.” <i>ArXiv</i>, n.d. <a href=\"https://doi.org/10.48550/arXiv.2205.11991\">https://doi.org/10.48550/arXiv.2205.11991</a>.","apa":"Zikelic, D., Lechner, M., Chatterjee, K., &#38; Henzinger, T. A. (n.d.). Learning stabilizing policies in stochastic control systems. <i>arXiv</i>. <a href=\"https://doi.org/10.48550/arXiv.2205.11991\">https://doi.org/10.48550/arXiv.2205.11991</a>","mla":"Zikelic, Dorde, et al. “Learning Stabilizing Policies in Stochastic Control Systems.” <i>ArXiv</i>, doi:<a href=\"https://doi.org/10.48550/arXiv.2205.11991\">10.48550/arXiv.2205.11991</a>."},"external_id":{"arxiv":["2205.11991"]},"related_material":{"record":[{"id":"14539","relation":"dissertation_contains","status":"public"}]},"arxiv":1,"month":"05","year":"2022","department":[{"_id":"KrCh"},{"_id":"ToHe"}],"abstract":[{"lang":"eng","text":"In this work, we address the problem of learning provably stable neural\r\nnetwork policies for stochastic control systems. While recent work has\r\ndemonstrated the feasibility of certifying given policies using martingale\r\ntheory, the problem of how to learn such policies is little explored. Here, we\r\nstudy the effectiveness of jointly learning a policy together with a martingale\r\ncertificate that proves its stability using a single learning algorithm. We\r\nobserve that the joint optimization problem becomes easily stuck in local\r\nminima when starting from a randomly initialized policy. Our results suggest\r\nthat some form of pre-training of the policy is required for the joint\r\noptimization to repair and verify the policy successfully."}],"publication_status":"submitted","main_file_link":[{"url":"https://arxiv.org/abs/2205.11991","open_access":"1"}],"oa_version":"Preprint","title":"Learning stabilizing policies in stochastic control systems","doi":"10.48550/arXiv.2205.11991","author":[{"orcid":"0000-0002-4681-1699","first_name":"Dorde","last_name":"Zikelic","full_name":"Zikelic, Dorde","id":"294AA7A6-F248-11E8-B48F-1D18A9856A87"},{"first_name":"Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","full_name":"Lechner, Mathias","last_name":"Lechner"},{"orcid":"0000-0002-4561-241X","first_name":"Krishnendu","last_name":"Chatterjee","id":"2E5DCA20-F248-11E8-B48F-1D18A9856A87","full_name":"Chatterjee, Krishnendu"},{"first_name":"Thomas A","orcid":"0000-0002-2985-7724","last_name":"Henzinger","full_name":"Henzinger, Thomas A","id":"40876CD8-F248-11E8-B48F-1D18A9856A87"}],"article_processing_charge":"No","day":"24","type":"preprint","date_created":"2023-11-24T13:22:30Z","date_updated":"2025-07-14T09:10:00Z","_id":"14601"},{"project":[{"name":"Vigilant Algorithmic Monitoring of Software","grant_number":"101020093","call_identifier":"H2020","_id":"62781420-2b32-11ec-9570-8d9b63373d4d"},{"_id":"25F42A32-B435-11E9-9278-68D0E5697425","grant_number":"Z211","name":"The Wittgenstein Prize","call_identifier":"FWF"}],"status":"public","publication":"2022 International Conference on Robotics and Automation","date_published":"2022-07-12T00:00:00Z","conference":{"location":"Philadelphia, PA, United States","name":"ICRA: International Conference on Robotics and Automation","end_date":"2022-05-27","start_date":"2022-05-23"},"acknowledgement":"L.B. was supported by the Doctoral College Resilient Embedded Systems. M.L. was supported in part by the ERC2020-AdG 101020093 and the Austrian Science Fund (FWF) under grant Z211-N23 (Wittgenstein Award). R.H. and D.R. were supported by The Boeing Company and the Office of Naval Research (ONR) Grant N00014-18-1-2830. R.G. was partially supported by the Horizon-2020 ECSEL Project grant No. 783163 (iDev40) and A.B. by FFG Project ADEX.","ec_funded":1,"external_id":{"arxiv":["2103.04909"]},"year":"2022","page":"7513-7520","quality_controlled":"1","main_file_link":[{"open_access":"1","url":"https://doi.org/10.48550/arXiv.2103.04909"}],"publisher":"IEEE","doi":"10.1109/ICRA46639.2022.9811650","article_processing_charge":"No","type":"conference","date_updated":"2022-09-05T08:46:12Z","_id":"12010","language":[{"iso":"eng"}],"oa":1,"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","citation":{"apa":"Brunnbauer, A., Berducci, L., Brandstatter, A., Lechner, M., Hasani, R., Rus, D., &#38; Grosu, R. (2022). Latent imagination facilitates zero-shot transfer in autonomous racing. In <i>2022 International Conference on Robotics and Automation</i> (pp. 7513–7520). Philadelphia, PA, United States: IEEE. <a href=\"https://doi.org/10.1109/ICRA46639.2022.9811650\">https://doi.org/10.1109/ICRA46639.2022.9811650</a>","mla":"Brunnbauer, Axel, et al. “Latent Imagination Facilitates Zero-Shot Transfer in Autonomous Racing.” <i>2022 International Conference on Robotics and Automation</i>, IEEE, 2022, pp. 7513–20, doi:<a href=\"https://doi.org/10.1109/ICRA46639.2022.9811650\">10.1109/ICRA46639.2022.9811650</a>.","ista":"Brunnbauer A, Berducci L, Brandstatter A, Lechner M, Hasani R, Rus D, Grosu R. 2022. Latent imagination facilitates zero-shot transfer in autonomous racing. 2022 International Conference on Robotics and Automation. ICRA: International Conference on Robotics and Automation, 7513–7520.","chicago":"Brunnbauer, Axel, Luigi Berducci, Andreas Brandstatter, Mathias Lechner, Ramin Hasani, Daniela Rus, and Radu Grosu. “Latent Imagination Facilitates Zero-Shot Transfer in Autonomous Racing.” In <i>2022 International Conference on Robotics and Automation</i>, 7513–20. IEEE, 2022. <a href=\"https://doi.org/10.1109/ICRA46639.2022.9811650\">https://doi.org/10.1109/ICRA46639.2022.9811650</a>.","ieee":"A. Brunnbauer <i>et al.</i>, “Latent imagination facilitates zero-shot transfer in autonomous racing,” in <i>2022 International Conference on Robotics and Automation</i>, Philadelphia, PA, United States, 2022, pp. 7513–7520.","short":"A. Brunnbauer, L. Berducci, A. Brandstatter, M. Lechner, R. Hasani, D. Rus, R. Grosu, in:, 2022 International Conference on Robotics and Automation, IEEE, 2022, pp. 7513–7520.","ama":"Brunnbauer A, Berducci L, Brandstatter A, et al. Latent imagination facilitates zero-shot transfer in autonomous racing. In: <i>2022 International Conference on Robotics and Automation</i>. IEEE; 2022:7513-7520. doi:<a href=\"https://doi.org/10.1109/ICRA46639.2022.9811650\">10.1109/ICRA46639.2022.9811650</a>"},"arxiv":1,"month":"07","department":[{"_id":"ToHe"}],"abstract":[{"text":"World models learn behaviors in a latent imagination space to enhance the sample-efficiency of deep reinforcement learning (RL) algorithms. While learning world models for high-dimensional observations (e.g., pixel inputs) has become practicable on standard RL benchmarks and some games, their effectiveness in real-world robotics applications has not been explored. In this paper, we investigate how such agents generalize to real-world autonomous vehicle control tasks, where advanced model-free deep RL algorithms fail. In particular, we set up a series of time-lap tasks for an F1TENTH racing robot, equipped with a high-dimensional LiDAR sensor, on a set of test tracks with a gradual increase in their complexity. In this continuous-control setting, we show that model-based agents capable of learning in imagination substantially outperform model-free agents with respect to performance, sample efficiency, successful task completion, and generalization. Moreover, we show that the generalization ability of model-based agents strongly depends on the choice of their observation model. We provide extensive empirical evidence for the effectiveness of world models provided with long enough memory horizons in sim2real tasks.","lang":"eng"}],"publication_status":"published","publication_identifier":{"issn":["1050-4729"],"isbn":["9781728196817"]},"oa_version":"Preprint","title":"Latent imagination facilitates zero-shot transfer in autonomous racing","author":[{"last_name":"Brunnbauer","full_name":"Brunnbauer, Axel","first_name":"Axel"},{"last_name":"Berducci","full_name":"Berducci, Luigi","first_name":"Luigi"},{"last_name":"Brandstatter","full_name":"Brandstatter, Andreas","first_name":"Andreas"},{"id":"3DC22916-F248-11E8-B48F-1D18A9856A87","full_name":"Lechner, Mathias","last_name":"Lechner","first_name":"Mathias"},{"full_name":"Hasani, Ramin","last_name":"Hasani","first_name":"Ramin"},{"first_name":"Daniela","last_name":"Rus","full_name":"Rus, Daniela"},{"first_name":"Radu","full_name":"Grosu, Radu","last_name":"Grosu"}],"day":"12","scopus_import":"1","date_created":"2022-09-04T22:02:02Z"},{"language":[{"iso":"eng"}],"oa":1,"user_id":"4359f0d1-fa6c-11eb-b949-802e58b17ae8","issue":"11","citation":{"ieee":"R. Hasani <i>et al.</i>, “Closed-form continuous-time neural networks,” <i>Nature Machine Intelligence</i>, vol. 4, no. 11. Springer Nature, pp. 992–1003, 2022.","short":"R. Hasani, M. Lechner, A. Amini, L. Liebenwein, A. Ray, M. Tschaikowski, G. Teschl, D. Rus, Nature Machine Intelligence 4 (2022) 992–1003.","ama":"Hasani R, Lechner M, Amini A, et al. Closed-form continuous-time neural networks. <i>Nature Machine Intelligence</i>. 2022;4(11):992-1003. doi:<a href=\"https://doi.org/10.1038/s42256-022-00556-7\">10.1038/s42256-022-00556-7</a>","apa":"Hasani, R., Lechner, M., Amini, A., Liebenwein, L., Ray, A., Tschaikowski, M., … Rus, D. (2022). Closed-form continuous-time neural networks. <i>Nature Machine Intelligence</i>. Springer Nature. <a href=\"https://doi.org/10.1038/s42256-022-00556-7\">https://doi.org/10.1038/s42256-022-00556-7</a>","mla":"Hasani, Ramin, et al. “Closed-Form Continuous-Time Neural Networks.” <i>Nature Machine Intelligence</i>, vol. 4, no. 11, Springer Nature, 2022, pp. 992–1003, doi:<a href=\"https://doi.org/10.1038/s42256-022-00556-7\">10.1038/s42256-022-00556-7</a>.","ista":"Hasani R, Lechner M, Amini A, Liebenwein L, Ray A, Tschaikowski M, Teschl G, Rus D. 2022. Closed-form continuous-time neural networks. Nature Machine Intelligence. 4(11), 992–1003.","chicago":"Hasani, Ramin, Mathias Lechner, Alexander Amini, Lucas Liebenwein, Aaron Ray, Max Tschaikowski, Gerald Teschl, and Daniela Rus. “Closed-Form Continuous-Time Neural Networks.” <i>Nature Machine Intelligence</i>. Springer Nature, 2022. <a href=\"https://doi.org/10.1038/s42256-022-00556-7\">https://doi.org/10.1038/s42256-022-00556-7</a>."},"arxiv":1,"month":"11","file":[{"date_created":"2023-01-24T09:49:44Z","file_size":3259553,"date_updated":"2023-01-24T09:49:44Z","creator":"dernst","file_id":"12355","file_name":"2022_NatureMachineIntelligence_Hasani.pdf","success":1,"content_type":"application/pdf","access_level":"open_access","relation":"main_file","checksum":"b4789122ce04bfb4ac042390f59aaa8b"}],"department":[{"_id":"ToHe"}],"tmp":{"name":"Creative Commons Attribution 4.0 International Public License (CC-BY 4.0)","legal_code_url":"https://creativecommons.org/licenses/by/4.0/legalcode","image":"/images/cc_by.png","short":"CC BY (4.0)"},"abstract":[{"lang":"eng","text":"Continuous-time neural networks are a class of machine learning systems that can tackle representation learning on spatiotemporal decision-making tasks. These models are typically represented by continuous differential equations. However, their expressive power when they are deployed on computers is bottlenecked by numerical differential equation solvers. This limitation has notably slowed down the scaling and understanding of numerous natural physical phenomena such as the dynamics of nervous systems. Ideally, we would circumvent this bottleneck by solving the given dynamical system in closed form. This is known to be intractable in general. Here, we show that it is possible to closely approximate the interaction between neurons and synapses—the building blocks of natural and artificial neural networks—constructed by liquid time-constant networks efficiently in closed form. To this end, we compute a tightly bounded approximation of the solution of an integral appearing in liquid time-constant dynamics that has had no known closed-form solution so far. This closed-form solution impacts the design of continuous-time and continuous-depth neural models. For instance, since time appears explicitly in closed form, the formulation relaxes the need for complex numerical solvers. Consequently, we obtain models that are between one and five orders of magnitude faster in training and inference compared with differential equation-based counterparts. More importantly, in contrast to ordinary differential equation-based continuous networks, closed-form networks can scale remarkably well compared with other deep learning instances. Lastly, as these models are derived from liquid networks, they show good performance in time-series modelling compared with advanced recurrent neural network models."}],"intvolume":"         4","has_accepted_license":"1","publication_identifier":{"issn":["2522-5839"]},"publication_status":"published","file_date_updated":"2023-01-24T09:49:44Z","oa_version":"Published Version","title":"Closed-form continuous-time neural networks","author":[{"full_name":"Hasani, Ramin","last_name":"Hasani","first_name":"Ramin"},{"first_name":"Mathias","full_name":"Lechner, Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","last_name":"Lechner"},{"first_name":"Alexander","last_name":"Amini","full_name":"Amini, Alexander"},{"full_name":"Liebenwein, Lucas","last_name":"Liebenwein","first_name":"Lucas"},{"full_name":"Ray, Aaron","last_name":"Ray","first_name":"Aaron"},{"first_name":"Max","last_name":"Tschaikowski","full_name":"Tschaikowski, Max"},{"last_name":"Teschl","full_name":"Teschl, Gerald","first_name":"Gerald"},{"full_name":"Rus, Daniela","last_name":"Rus","first_name":"Daniela"}],"day":"15","scopus_import":"1","article_type":"original","date_created":"2023-01-12T12:07:21Z","volume":4,"project":[{"_id":"25F42A32-B435-11E9-9278-68D0E5697425","name":"The Wittgenstein Prize","grant_number":"Z211","call_identifier":"FWF"}],"status":"public","publication":"Nature Machine Intelligence","date_published":"2022-11-15T00:00:00Z","acknowledgement":"This research was supported in part by the AI2050 program at Schmidt Futures (grant G-22-63172), the Boeing Company, and the United States Air Force Research Laboratory and the United States Air Force Artificial Intelligence Accelerator and was accomplished under cooperative agreement number FA8750-19-2-1000. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the United States Air Force or the U.S. Government. The U.S. Government is authorized to reproduce and distribute reprints for Government purposes, notwithstanding any copyright notation herein. This work was further supported by The Boeing Company and Office of Naval Research grant N00014-18-1-2830. M.T. is supported by the Poul Due Jensen Foundation, grant 883901. M.L. was supported in part by the Austrian Science Fund under grant Z211-N23 (Wittgenstein Award). A.A. was supported by the National Science Foundation Graduate Research Fellowship Program. We thank T.-H. Wang, P. Kao, M. Chahine, W. Xiao, X. Li, L. Yin and Y. Ben for useful suggestions and for testing of CfC models to confirm the results across other domains.","external_id":{"isi":["000884215600003"],"arxiv":["2106.13898"]},"related_material":{"link":[{"url":"https://doi.org/10.1038/s42256-022-00597-y","relation":"erratum"}]},"year":"2022","isi":1,"keyword":["Artificial Intelligence","Computer Networks and Communications","Computer Vision and Pattern Recognition","Human-Computer Interaction","Software"],"ddc":["000"],"page":"992-1003","quality_controlled":"1","publisher":"Springer Nature","doi":"10.1038/s42256-022-00556-7","article_processing_charge":"No","type":"journal_article","date_updated":"2023-08-04T09:00:10Z","_id":"12147"},{"publication_identifier":{"eissn":["2374-3468"],"issn":["2159-5399"],"isbn":["978577358350"]},"publication_status":"published","intvolume":"        36","abstract":[{"text":"We introduce a new statistical verification algorithm that formally quantifies the behavioral robustness of any time-continuous process formulated as a continuous-depth model. Our algorithm solves a set of global optimization (Go) problems over a given time horizon to construct a tight enclosure (Tube) of the set of all process executions starting from a ball of initial states. We call our algorithm GoTube. Through its construction, GoTube ensures that the bounding tube is conservative up to a desired probability and up to a desired tightness.\r\n GoTube is implemented in JAX and optimized to scale to complex continuous-depth neural network models. Compared to advanced reachability analysis tools for time-continuous neural networks, GoTube does not accumulate overapproximation errors between time steps and avoids the infamous wrapping effect inherent in symbolic techniques. We show that GoTube substantially outperforms state-of-the-art verification tools in terms of the size of the initial ball, speed, time-horizon, task completion, and scalability on a large set of experiments.\r\n GoTube is stable and sets the state-of-the-art in terms of its ability to scale to time horizons well beyond what has been previously possible.","lang":"eng"}],"article_type":"original","date_created":"2023-02-05T17:27:42Z","volume":36,"title":"GoTube: Scalable statistical verification of continuous-depth models","oa_version":"Preprint","author":[{"last_name":"Gruenbacher","full_name":"Gruenbacher, Sophie A.","first_name":"Sophie A."},{"full_name":"Lechner, Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","last_name":"Lechner","first_name":"Mathias"},{"last_name":"Hasani","full_name":"Hasani, Ramin","first_name":"Ramin"},{"full_name":"Rus, Daniela","last_name":"Rus","first_name":"Daniela"},{"orcid":"0000-0002-2985-7724","first_name":"Thomas A","last_name":"Henzinger","id":"40876CD8-F248-11E8-B48F-1D18A9856A87","full_name":"Henzinger, Thomas A"},{"full_name":"Smolka, Scott A.","last_name":"Smolka","first_name":"Scott A."},{"first_name":"Radu","full_name":"Grosu, Radu","last_name":"Grosu"}],"day":"28","scopus_import":"1","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","issue":"6","citation":{"apa":"Gruenbacher, S. A., Lechner, M., Hasani, R., Rus, D., Henzinger, T. A., Smolka, S. A., &#38; Grosu, R. (2022). GoTube: Scalable statistical verification of continuous-depth models. <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>. Association for the Advancement of Artificial Intelligence. <a href=\"https://doi.org/10.1609/aaai.v36i6.20631\">https://doi.org/10.1609/aaai.v36i6.20631</a>","mla":"Gruenbacher, Sophie A., et al. “GoTube: Scalable Statistical Verification of Continuous-Depth Models.” <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>, vol. 36, no. 6, Association for the Advancement of Artificial Intelligence, 2022, pp. 6755–64, doi:<a href=\"https://doi.org/10.1609/aaai.v36i6.20631\">10.1609/aaai.v36i6.20631</a>.","ista":"Gruenbacher SA, Lechner M, Hasani R, Rus D, Henzinger TA, Smolka SA, Grosu R. 2022. GoTube: Scalable statistical verification of continuous-depth models. Proceedings of the AAAI Conference on Artificial Intelligence. 36(6), 6755–6764.","chicago":"Gruenbacher, Sophie A., Mathias Lechner, Ramin Hasani, Daniela Rus, Thomas A Henzinger, Scott A. Smolka, and Radu Grosu. “GoTube: Scalable Statistical Verification of Continuous-Depth Models.” <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>. Association for the Advancement of Artificial Intelligence, 2022. <a href=\"https://doi.org/10.1609/aaai.v36i6.20631\">https://doi.org/10.1609/aaai.v36i6.20631</a>.","ieee":"S. A. Gruenbacher <i>et al.</i>, “GoTube: Scalable statistical verification of continuous-depth models,” <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>, vol. 36, no. 6. Association for the Advancement of Artificial Intelligence, pp. 6755–6764, 2022.","short":"S.A. Gruenbacher, M. Lechner, R. Hasani, D. Rus, T.A. Henzinger, S.A. Smolka, R. Grosu, Proceedings of the AAAI Conference on Artificial Intelligence 36 (2022) 6755–6764.","ama":"Gruenbacher SA, Lechner M, Hasani R, et al. GoTube: Scalable statistical verification of continuous-depth models. <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>. 2022;36(6):6755-6764. doi:<a href=\"https://doi.org/10.1609/aaai.v36i6.20631\">10.1609/aaai.v36i6.20631</a>"},"language":[{"iso":"eng"}],"oa":1,"department":[{"_id":"ToHe"}],"arxiv":1,"month":"06","quality_controlled":"1","main_file_link":[{"open_access":"1","url":"https://arxiv.org/abs/2107.08467"}],"page":"6755-6764","type":"journal_article","date_updated":"2023-09-26T10:46:59Z","_id":"12510","publisher":"Association for the Advancement of Artificial Intelligence","doi":"10.1609/aaai.v36i6.20631","article_processing_charge":"No","acknowledgement":"SG is funded by the Austrian Science Fund (FWF) project number W1255-N23. ML and TH are supported in part by FWF under grant Z211-N23 (Wittgenstein Award) and the ERC-2020-AdG 101020093. SS is supported by NSF awards DCL-2040599, CCF-1918225, and CPS-1446832. RH and DR are partially supported by Boeing. RG is partially supported by Horizon-2020 ECSEL Project grant No. 783163 (iDev40).","date_published":"2022-06-28T00:00:00Z","ec_funded":1,"project":[{"_id":"25F42A32-B435-11E9-9278-68D0E5697425","grant_number":"Z211","name":"The Wittgenstein Prize","call_identifier":"FWF"},{"_id":"62781420-2b32-11ec-9570-8d9b63373d4d","call_identifier":"H2020","name":"Vigilant Algorithmic Monitoring of Software","grant_number":"101020093"}],"publication":"Proceedings of the AAAI Conference on Artificial Intelligence","status":"public","keyword":["General Medicine"],"external_id":{"arxiv":["2107.08467"]},"year":"2022"},{"abstract":[{"lang":"eng","text":"We consider the problem of formally verifying almost-sure (a.s.) asymptotic stability in discrete-time nonlinear stochastic control systems. While verifying stability in deterministic control systems is extensively studied in the literature, verifying stability in stochastic control systems is an open problem. The few existing works on this topic either consider only specialized forms of stochasticity or make restrictive assumptions on the system, rendering them inapplicable to learning algorithms with neural network policies. \r\n In this work, we present an approach for general nonlinear stochastic control problems with two novel aspects: (a) instead of classical stochastic extensions of Lyapunov functions, we use ranking supermartingales (RSMs) to certify a.s. asymptotic stability, and (b) we present a method for learning neural network RSMs. \r\n We prove that our approach guarantees a.s. asymptotic stability of the system and\r\n provides the first method to obtain bounds on the stabilization time, which stochastic Lyapunov functions do not.\r\n Finally, we validate our approach experimentally on a set of nonlinear stochastic reinforcement learning environments with neural network policies."}],"intvolume":"        36","publication_identifier":{"isbn":["9781577358350"],"issn":["2159-5399"],"eissn":["2374-3468"]},"publication_status":"published","title":"Stability verification in stochastic control systems via neural network supermartingales","oa_version":"Preprint","day":"28","scopus_import":"1","author":[{"last_name":"Lechner","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","full_name":"Lechner, Mathias","first_name":"Mathias"},{"orcid":"0000-0002-4681-1699","first_name":"Dorde","id":"294AA7A6-F248-11E8-B48F-1D18A9856A87","full_name":"Zikelic, Dorde","last_name":"Zikelic"},{"id":"2E5DCA20-F248-11E8-B48F-1D18A9856A87","full_name":"Chatterjee, Krishnendu","last_name":"Chatterjee","orcid":"0000-0002-4561-241X","first_name":"Krishnendu"},{"orcid":"0000-0002-2985-7724","first_name":"Thomas A","last_name":"Henzinger","id":"40876CD8-F248-11E8-B48F-1D18A9856A87","full_name":"Henzinger, Thomas A"}],"date_created":"2023-02-05T17:29:50Z","article_type":"original","volume":36,"language":[{"iso":"eng"}],"oa":1,"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","citation":{"chicago":"Lechner, Mathias, Dorde Zikelic, Krishnendu Chatterjee, and Thomas A Henzinger. “Stability Verification in Stochastic Control Systems via Neural Network Supermartingales.” <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>. Association for the Advancement of Artificial Intelligence, 2022. <a href=\"https://doi.org/10.1609/aaai.v36i7.20695\">https://doi.org/10.1609/aaai.v36i7.20695</a>.","ista":"Lechner M, Zikelic D, Chatterjee K, Henzinger TA. 2022. Stability verification in stochastic control systems via neural network supermartingales. Proceedings of the AAAI Conference on Artificial Intelligence. 36(7), 7326–7336.","apa":"Lechner, M., Zikelic, D., Chatterjee, K., &#38; Henzinger, T. A. (2022). Stability verification in stochastic control systems via neural network supermartingales. <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>. Association for the Advancement of Artificial Intelligence. <a href=\"https://doi.org/10.1609/aaai.v36i7.20695\">https://doi.org/10.1609/aaai.v36i7.20695</a>","mla":"Lechner, Mathias, et al. “Stability Verification in Stochastic Control Systems via Neural Network Supermartingales.” <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>, vol. 36, no. 7, Association for the Advancement of Artificial Intelligence, 2022, pp. 7326–36, doi:<a href=\"https://doi.org/10.1609/aaai.v36i7.20695\">10.1609/aaai.v36i7.20695</a>.","ama":"Lechner M, Zikelic D, Chatterjee K, Henzinger TA. Stability verification in stochastic control systems via neural network supermartingales. <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>. 2022;36(7):7326-7336. doi:<a href=\"https://doi.org/10.1609/aaai.v36i7.20695\">10.1609/aaai.v36i7.20695</a>","ieee":"M. Lechner, D. Zikelic, K. Chatterjee, and T. A. Henzinger, “Stability verification in stochastic control systems via neural network supermartingales,” <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>, vol. 36, no. 7. Association for the Advancement of Artificial Intelligence, pp. 7326–7336, 2022.","short":"M. Lechner, D. Zikelic, K. Chatterjee, T.A. Henzinger, Proceedings of the AAAI Conference on Artificial Intelligence 36 (2022) 7326–7336."},"issue":"7","month":"06","arxiv":1,"department":[{"_id":"ToHe"},{"_id":"KrCh"}],"page":"7326-7336","quality_controlled":"1","main_file_link":[{"open_access":"1","url":"https://arxiv.org/abs/2112.09495"}],"publisher":"Association for the Advancement of Artificial Intelligence","article_processing_charge":"No","doi":"10.1609/aaai.v36i7.20695","type":"journal_article","_id":"12511","date_updated":"2025-07-14T09:09:58Z","status":"public","publication":"Proceedings of the AAAI Conference on Artificial Intelligence","project":[{"call_identifier":"H2020","name":"Vigilant Algorithmic Monitoring of Software","grant_number":"101020093","_id":"62781420-2b32-11ec-9570-8d9b63373d4d"},{"_id":"0599E47C-7A3F-11EA-A408-12923DDC885E","call_identifier":"H2020","name":"Formal Methods for Stochastic Models: Algorithms and Applications","grant_number":"863818"},{"name":"International IST Doctoral Program","grant_number":"665385","call_identifier":"H2020","_id":"2564DBCA-B435-11E9-9278-68D0E5697425"}],"date_published":"2022-06-28T00:00:00Z","acknowledgement":"This work was supported in part by the ERC-2020-AdG 101020093, ERC CoG 863818 (FoRM-SMArt) and the European Union’s Horizon 2020 research and innovation programme\r\nunder the Marie Skłodowska-Curie Grant Agreement No. 665385.","ec_funded":1,"related_material":{"record":[{"id":"14539","relation":"dissertation_contains","status":"public"}]},"external_id":{"arxiv":["2112.09495"]},"year":"2022","keyword":["General Medicine"]},{"type":"conference","date_updated":"2025-07-14T09:10:11Z","_id":"10665","publisher":"AAAI Press","article_processing_charge":"No","alternative_title":["Technical Tracks"],"quality_controlled":"1","main_file_link":[{"open_access":"1","url":"https://ojs.aaai.org/index.php/AAAI/article/view/16496"}],"ddc":["000"],"page":"3787-3795","external_id":{"arxiv":["2012.08185"]},"related_material":{"record":[{"status":"public","relation":"dissertation_contains","id":"11362"}]},"year":"2021","date_published":"2021-05-28T00:00:00Z","acknowledgement":"This research was supported in part by the Austrian Science Fund (FWF) under grant Z211-N23 (Wittgenstein\r\nAward), 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.\r\n","conference":{"location":"Virtual","start_date":"2021-02-02","end_date":"2021-02-09","name":"AAAI: Association for the Advancement of Artificial Intelligence"},"ec_funded":1,"project":[{"grant_number":"665385","name":"International IST Doctoral Program","call_identifier":"H2020","_id":"2564DBCA-B435-11E9-9278-68D0E5697425"},{"_id":"25F42A32-B435-11E9-9278-68D0E5697425","grant_number":"Z211","name":"The Wittgenstein Prize","call_identifier":"FWF"},{"_id":"0599E47C-7A3F-11EA-A408-12923DDC885E","grant_number":"863818","name":"Formal Methods for Stochastic Models: Algorithms and Applications","call_identifier":"H2020"}],"status":"public","publication":"Proceedings of the AAAI Conference on Artificial Intelligence","date_created":"2022-01-25T15:15:02Z","volume":35,"title":"Scalable verification of quantized neural networks","oa_version":"Published Version","author":[{"first_name":"Thomas A","orcid":"0000-0002-2985-7724","full_name":"Henzinger, Thomas A","id":"40876CD8-F248-11E8-B48F-1D18A9856A87","last_name":"Henzinger"},{"first_name":"Mathias","full_name":"Lechner, Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","last_name":"Lechner"},{"first_name":"Dorde","orcid":"0000-0002-4681-1699","last_name":"Zikelic","id":"294AA7A6-F248-11E8-B48F-1D18A9856A87","full_name":"Zikelic, Dorde"}],"day":"28","scopus_import":"1","publication_status":"published","publication_identifier":{"isbn":["978-1-57735-866-4"],"issn":["2159-5399"],"eissn":["2374-3468"]},"file_date_updated":"2022-01-26T07:41:16Z","abstract":[{"text":"Formal verification of neural networks is an active topic of research, and recent advances have significantly increased the size of the networks that verification tools can handle. However, most methods are designed for verification of an idealized model of the actual network which works over real arithmetic and ignores rounding imprecisions. This idealization is in stark contrast to network quantization, which is a technique that trades numerical precision for computational efficiency and is, therefore, often applied in practice. Neglecting rounding errors of such low-bit quantized neural networks has been shown to lead to wrong conclusions about the network’s correctness. Thus, the desired approach for verifying quantized neural networks would be one that takes these rounding errors\r\ninto account. In this paper, we show that verifying the bitexact implementation of quantized neural networks with bitvector specifications is PSPACE-hard, even though verifying idealized real-valued networks and satisfiability of bit-vector specifications alone are each in NP. Furthermore, we explore several practical heuristics toward closing the complexity gap between idealized and bit-exact verification. In particular, we propose three techniques for making SMT-based verification of quantized neural networks more scalable. Our experiments demonstrate that our proposed methods allow a speedup of up to three orders of magnitude over existing approaches.","lang":"eng"}],"intvolume":"        35","has_accepted_license":"1","file":[{"file_id":"10684","date_created":"2022-01-26T07:41:16Z","file_size":137235,"date_updated":"2022-01-26T07:41:16Z","creator":"mlechner","relation":"main_file","checksum":"2bc8155b2526a70fba5b7301bc89dbd1","success":1,"file_name":"16496-Article Text-19990-1-2-20210518 (1).pdf","access_level":"open_access","content_type":"application/pdf"}],"department":[{"_id":"GradSch"},{"_id":"ToHe"}],"month":"05","arxiv":1,"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","issue":"5A","citation":{"ama":"Henzinger TA, Lechner M, Zikelic D. Scalable verification of quantized neural networks. In: <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>. Vol 35. AAAI Press; 2021:3787-3795.","short":"T.A. Henzinger, M. Lechner, D. Zikelic, in:, Proceedings of the AAAI Conference on Artificial Intelligence, AAAI Press, 2021, pp. 3787–3795.","ieee":"T. A. Henzinger, M. Lechner, and D. Zikelic, “Scalable verification of quantized neural networks,” in <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>, Virtual, 2021, vol. 35, no. 5A, pp. 3787–3795.","chicago":"Henzinger, Thomas A, Mathias Lechner, and Dorde Zikelic. “Scalable Verification of Quantized Neural Networks.” In <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>, 35:3787–95. AAAI Press, 2021.","ista":"Henzinger TA, Lechner M, Zikelic D. 2021. Scalable verification of quantized neural networks. Proceedings of the AAAI Conference on Artificial Intelligence. AAAI: Association for the Advancement of Artificial Intelligence, Technical Tracks, vol. 35, 3787–3795.","mla":"Henzinger, Thomas A., et al. “Scalable Verification of Quantized Neural Networks.” <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>, vol. 35, no. 5A, AAAI Press, 2021, pp. 3787–95.","apa":"Henzinger, T. A., Lechner, M., &#38; Zikelic, D. (2021). Scalable verification of quantized neural networks. In <i>Proceedings of the AAAI Conference on Artificial Intelligence</i> (Vol. 35, pp. 3787–3795). Virtual: AAAI Press."},"language":[{"iso":"eng"}],"oa":1},{"date_created":"2022-01-25T15:44:54Z","author":[{"last_name":"Lechner","full_name":"Lechner, Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","first_name":"Mathias"},{"first_name":"Ramin","last_name":"Hasani","full_name":"Hasani, Ramin"},{"first_name":"Radu","full_name":"Grosu, Radu","last_name":"Grosu"},{"first_name":"Daniela","full_name":"Rus, Daniela","last_name":"Rus"},{"id":"40876CD8-F248-11E8-B48F-1D18A9856A87","full_name":"Henzinger, Thomas A","last_name":"Henzinger","orcid":"0000-0002-2985-7724","first_name":"Thomas A"}],"title":"Adversarial training is not ready for robot learning","oa_version":"None","publication_identifier":{"isbn":["978-1-7281-9078-5"],"issn":["1050-4729"],"eisbn":["978-1-7281-9077-8"],"eissn":["2577-087X"]},"publication_status":"published","has_accepted_license":"1","abstract":[{"lang":"eng","text":"Adversarial training is an effective method to train deep learning models that are resilient to norm-bounded perturbations, with the cost of nominal performance drop. While adversarial training appears to enhance the robustness and safety of a deep model deployed in open-world decision-critical applications, counterintuitively, it induces undesired behaviors in robot learning settings. In this paper, we show theoretically and experimentally that neural controllers obtained via adversarial training are subjected to three types of defects, namely transient, systematic, and conditional errors. We first generalize adversarial training to a safety-domain optimization scheme allowing for more generic specifications. We then prove that such a learning process tends to cause certain error profiles. We support our theoretical results by a thorough experimental safety analysis in a robot-learning task. Our results suggest that adversarial training is not yet ready for robot learning."}],"license":"https://creativecommons.org/licenses/by-nc-nd/3.0/","tmp":{"image":"/images/cc_by_nc_nd.png","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","short":"CC BY-NC-ND (3.0)"},"department":[{"_id":"GradSch"},{"_id":"ToHe"}],"arxiv":1,"citation":{"ama":"Lechner M, Hasani R, Grosu R, Rus D, Henzinger TA. Adversarial training is not ready for robot learning. In: <i>2021 IEEE International Conference on Robotics and Automation</i>. ICRA. ; 2021:4140-4147. doi:<a href=\"https://doi.org/10.1109/ICRA48506.2021.9561036\">10.1109/ICRA48506.2021.9561036</a>","ieee":"M. Lechner, R. Hasani, R. Grosu, D. Rus, and T. A. Henzinger, “Adversarial training is not ready for robot learning,” in <i>2021 IEEE International Conference on Robotics and Automation</i>, Xi’an, China, 2021, pp. 4140–4147.","short":"M. Lechner, R. Hasani, R. Grosu, D. Rus, T.A. Henzinger, in:, 2021 IEEE International Conference on Robotics and Automation, 2021, pp. 4140–4147.","chicago":"Lechner, Mathias, Ramin Hasani, Radu Grosu, Daniela Rus, and Thomas A Henzinger. “Adversarial Training Is Not Ready for Robot Learning.” In <i>2021 IEEE International Conference on Robotics and Automation</i>, 4140–47. ICRA, 2021. <a href=\"https://doi.org/10.1109/ICRA48506.2021.9561036\">https://doi.org/10.1109/ICRA48506.2021.9561036</a>.","ista":"Lechner M, Hasani R, Grosu R, Rus D, Henzinger TA. 2021. Adversarial training is not ready for robot learning. 2021 IEEE International Conference on Robotics and Automation. ICRA: International Conference on Robotics and AutomationICRA, 4140–4147.","apa":"Lechner, M., Hasani, R., Grosu, R., Rus, D., &#38; Henzinger, T. A. (2021). Adversarial training is not ready for robot learning. In <i>2021 IEEE International Conference on Robotics and Automation</i> (pp. 4140–4147). Xi’an, China. <a href=\"https://doi.org/10.1109/ICRA48506.2021.9561036\">https://doi.org/10.1109/ICRA48506.2021.9561036</a>","mla":"Lechner, Mathias, et al. “Adversarial Training Is Not Ready for Robot Learning.” <i>2021 IEEE International Conference on Robotics and Automation</i>, 2021, pp. 4140–47, doi:<a href=\"https://doi.org/10.1109/ICRA48506.2021.9561036\">10.1109/ICRA48506.2021.9561036</a>."},"user_id":"4359f0d1-fa6c-11eb-b949-802e58b17ae8","oa":1,"language":[{"iso":"eng"}],"date_updated":"2023-08-17T06:58:38Z","_id":"10666","type":"conference","series_title":"ICRA","doi":"10.1109/ICRA48506.2021.9561036","article_processing_charge":"No","main_file_link":[{"url":"https://arxiv.org/abs/2103.08187","open_access":"1"}],"quality_controlled":"1","page":"4140-4147","ddc":["000"],"isi":1,"year":"2021","external_id":{"arxiv":["2103.08187"],"isi":["000765738803040"]},"related_material":{"record":[{"id":"11362","status":"public","relation":"dissertation_contains"}]},"acknowledgement":"M.L. and T.A.H. are supported in part by the Austrian Science Fund (FWF) under grant Z211-N23 (Wittgenstein Award). R.H. and D.R. are supported by Boeing and R.G. by Horizon-2020 ECSEL Project grant no. 783163 (iDev40).","date_published":"2021-01-01T00:00:00Z","conference":{"location":"Xi'an, China","name":"ICRA: International Conference on Robotics and Automation","start_date":"2021-05-30","end_date":"2021-06-05"},"project":[{"_id":"25F42A32-B435-11E9-9278-68D0E5697425","name":"The Wittgenstein Prize","grant_number":"Z211","call_identifier":"FWF"}],"publication":"2021 IEEE International Conference on Robotics and Automation","status":"public"},{"language":[{"iso":"eng"}],"oa":1,"user_id":"2EBD1598-F248-11E8-B48F-1D18A9856A87","citation":{"mla":"Lechner, Mathias, et al. “Infinite Time Horizon Safety of Bayesian Neural Networks.” <i>35th Conference on Neural Information Processing Systems</i>, 2021, doi:<a href=\"https://doi.org/10.48550/arXiv.2111.03165\">10.48550/arXiv.2111.03165</a>.","apa":"Lechner, M., Žikelić, Ð., Chatterjee, K., &#38; Henzinger, T. A. (2021). Infinite time horizon safety of Bayesian neural networks. In <i>35th Conference on Neural Information Processing Systems</i>. Virtual. <a href=\"https://doi.org/10.48550/arXiv.2111.03165\">https://doi.org/10.48550/arXiv.2111.03165</a>","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, .","chicago":"Lechner, Mathias, Ðorđe Žikelić, Krishnendu Chatterjee, and Thomas A Henzinger. “Infinite Time Horizon Safety of Bayesian Neural Networks.” In <i>35th Conference on Neural Information Processing Systems</i>, 2021. <a href=\"https://doi.org/10.48550/arXiv.2111.03165\">https://doi.org/10.48550/arXiv.2111.03165</a>.","short":"M. Lechner, Ð. Žikelić, K. Chatterjee, T.A. Henzinger, in:, 35th Conference on Neural Information Processing Systems, 2021.","ieee":"M. Lechner, Ð. Žikelić, K. Chatterjee, and T. A. Henzinger, “Infinite time horizon safety of Bayesian neural networks,” in <i>35th Conference on Neural Information Processing Systems</i>, Virtual, 2021.","ama":"Lechner M, Žikelić Ð, Chatterjee K, Henzinger TA. Infinite time horizon safety of Bayesian neural networks. In: <i>35th Conference on Neural Information Processing Systems</i>. ; 2021. doi:<a href=\"https://doi.org/10.48550/arXiv.2111.03165\">10.48550/arXiv.2111.03165</a>"},"month":"12","arxiv":1,"file":[{"content_type":"application/pdf","access_level":"open_access","file_name":"infinite_time_horizon_safety_o.pdf","success":1,"checksum":"0fc0f852525c10dda9cc9ffea07fb4e4","relation":"main_file","date_updated":"2022-01-26T07:39:59Z","creator":"mlechner","date_created":"2022-01-26T07:39:59Z","file_size":452492,"file_id":"10682"}],"department":[{"_id":"GradSch"},{"_id":"ToHe"},{"_id":"KrCh"}],"tmp":{"image":"/images/cc_by_nc_nd.png","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","short":"CC BY-NC-ND (3.0)"},"abstract":[{"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.","lang":"eng"}],"has_accepted_license":"1","publication_status":"published","file_date_updated":"2022-01-26T07:39:59Z","title":"Infinite time horizon safety of Bayesian neural networks","oa_version":"Published Version","author":[{"first_name":"Mathias","full_name":"Lechner, Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","last_name":"Lechner"},{"first_name":"Ðorđe","full_name":"Žikelić, Ðorđe","last_name":"Žikelić"},{"full_name":"Chatterjee, Krishnendu","id":"2E5DCA20-F248-11E8-B48F-1D18A9856A87","last_name":"Chatterjee","first_name":"Krishnendu","orcid":"0000-0002-4561-241X"},{"orcid":"0000-0002-2985-7724","first_name":"Thomas A","full_name":"Henzinger, Thomas A","id":"40876CD8-F248-11E8-B48F-1D18A9856A87","last_name":"Henzinger"}],"day":"01","date_created":"2022-01-25T15:45:58Z","project":[{"_id":"2564DBCA-B435-11E9-9278-68D0E5697425","grant_number":"665385","name":"International IST Doctoral Program","call_identifier":"H2020"},{"_id":"0599E47C-7A3F-11EA-A408-12923DDC885E","call_identifier":"H2020","grant_number":"863818","name":"Formal Methods for Stochastic Models: Algorithms and Applications"},{"grant_number":"Z211","name":"The Wittgenstein Prize","call_identifier":"FWF","_id":"25F42A32-B435-11E9-9278-68D0E5697425"}],"publication":"35th Conference on Neural Information Processing Systems","status":"public","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_published":"2021-12-01T00:00:00Z","conference":{"name":"NeurIPS: Neural Information Processing Systems","end_date":"2021-12-10","start_date":"2021-12-06","location":"Virtual"},"ec_funded":1,"external_id":{"arxiv":["2111.03165"]},"related_material":{"record":[{"status":"public","relation":"dissertation_contains","id":"11362"}]},"year":"2021","ddc":["000"],"quality_controlled":"1","main_file_link":[{"open_access":"1","url":"https://proceedings.neurips.cc/paper/2021/hash/544defa9fddff50c53b71c43e0da72be-Abstract.html"}],"doi":"10.48550/arXiv.2111.03165","alternative_title":[" Advances in Neural Information Processing Systems"],"article_processing_charge":"No","type":"conference","date_updated":"2025-07-14T09:10:12Z","_id":"10667"},{"volume":139,"date_created":"2022-01-25T15:46:33Z","day":"01","author":[{"first_name":"Zahra","full_name":"Babaiee, Zahra","last_name":"Babaiee"},{"first_name":"Ramin","full_name":"Hasani, Ramin","last_name":"Hasani"},{"first_name":"Mathias","last_name":"Lechner","full_name":"Lechner, Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87"},{"first_name":"Daniela","full_name":"Rus, Daniela","last_name":"Rus"},{"first_name":"Radu","full_name":"Grosu, Radu","last_name":"Grosu"}],"title":"On-off center-surround receptive fields for accurate and robust image classification","oa_version":"Published Version","file_date_updated":"2022-01-26T07:38:32Z","publication_identifier":{"issn":["2640-3498"]},"publication_status":"published","has_accepted_license":"1","tmp":{"image":"/images/cc_by_nc_nd.png","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","short":"CC BY-NC-ND (3.0)"},"intvolume":"       139","abstract":[{"lang":"eng","text":"Robustness to variations in lighting conditions is a key objective for any deep vision system. To this end, our paper extends the receptive field of convolutional neural networks with two residual components, ubiquitous in the visual processing system of vertebrates: On-center and off-center pathways, with an excitatory center and inhibitory surround; OOCS for short. The On-center pathway is excited by the presence of a light stimulus in its center, but not in its surround, whereas the Off-center pathway is excited by the absence of a light stimulus in its center, but not in its surround. We design OOCS pathways via a difference of Gaussians, with their variance computed analytically from the size of the receptive fields. OOCS pathways complement each other in their response to light stimuli, ensuring this way a strong edge-detection capability, and as a result an accurate and robust inference under challenging lighting conditions. We provide extensive empirical evidence showing that networks supplied with OOCS pathways gain accuracy and illumination-robustness from the novel edge representation, compared to other baselines."}],"department":[{"_id":"GradSch"},{"_id":"ToHe"}],"file":[{"creator":"mlechner","date_updated":"2022-01-26T07:38:32Z","file_size":4246561,"date_created":"2022-01-26T07:38:32Z","file_id":"10681","content_type":"application/pdf","access_level":"open_access","file_name":"babaiee21a.pdf","success":1,"checksum":"d30eae62561bb517d9f978437d7677db","relation":"main_file"}],"month":"07","citation":{"ama":"Babaiee Z, Hasani R, Lechner M, Rus D, Grosu R. On-off center-surround receptive fields for accurate and robust image classification. In: <i>Proceedings of the 38th International Conference on Machine Learning</i>. Vol 139. ML Research Press; 2021:478-489.","ieee":"Z. Babaiee, R. Hasani, M. Lechner, D. Rus, and R. Grosu, “On-off center-surround receptive fields for accurate and robust image classification,” in <i>Proceedings of the 38th International Conference on Machine Learning</i>, Virtual, 2021, vol. 139, pp. 478–489.","short":"Z. Babaiee, R. Hasani, M. Lechner, D. Rus, R. Grosu, in:, Proceedings of the 38th International Conference on Machine Learning, ML Research Press, 2021, pp. 478–489.","ista":"Babaiee Z, Hasani R, Lechner M, Rus D, Grosu R. 2021. On-off center-surround receptive fields for accurate and robust image classification. Proceedings of the 38th International Conference on Machine Learning. ML: Machine Learning, PMLR, vol. 139, 478–489.","chicago":"Babaiee, Zahra, Ramin Hasani, Mathias Lechner, Daniela Rus, and Radu Grosu. “On-off Center-Surround Receptive Fields for Accurate and Robust Image Classification.” In <i>Proceedings of the 38th International Conference on Machine Learning</i>, 139:478–89. ML Research Press, 2021.","apa":"Babaiee, Z., Hasani, R., Lechner, M., Rus, D., &#38; Grosu, R. (2021). On-off center-surround receptive fields for accurate and robust image classification. In <i>Proceedings of the 38th International Conference on Machine Learning</i> (Vol. 139, pp. 478–489). Virtual: ML Research Press.","mla":"Babaiee, Zahra, et al. “On-off Center-Surround Receptive Fields for Accurate and Robust Image Classification.” <i>Proceedings of the 38th International Conference on Machine Learning</i>, vol. 139, ML Research Press, 2021, pp. 478–89."},"user_id":"2EBD1598-F248-11E8-B48F-1D18A9856A87","oa":1,"language":[{"iso":"eng"}],"_id":"10668","date_updated":"2022-05-04T15:02:27Z","type":"conference","article_processing_charge":"No","alternative_title":["PMLR"],"publisher":"ML Research Press","main_file_link":[{"open_access":"1","url":"https://proceedings.mlr.press/v139/babaiee21a"}],"quality_controlled":"1","page":"478-489","ddc":["000"],"year":"2021","acknowledgement":"Z.B. is supported by the Doctoral College Resilient Embedded Systems, which is run jointly by the TU Wien’s Faculty of Informatics and the UAS Technikum Wien. R.G. is partially supported by the Horizon 2020 Era-Permed project Persorad, and ECSEL Project grant no. 783163 (iDev40). R.H and D.R were partially supported by Boeing and MIT. M.L. is supported in part by the Austrian Science Fund (FWF) under grant Z211-N23 (Wittgenstein Award).","conference":{"location":"Virtual","name":"ML: Machine Learning","end_date":"2021-07-24","start_date":"2021-07-18"},"date_published":"2021-07-01T00:00:00Z","status":"public","publication":"Proceedings of the 38th International Conference on Machine Learning","project":[{"_id":"25F42A32-B435-11E9-9278-68D0E5697425","call_identifier":"FWF","grant_number":"Z211","name":"The Wittgenstein Prize"}]},{"language":[{"iso":"eng"}],"oa":1,"user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","citation":{"ista":"Grunbacher S, Hasani R, Lechner M, Cyranka J, Smolka SA, Grosu R. 2021. On the verification of neural ODEs with stochastic guarantees. Proceedings of the AAAI Conference on Artificial Intelligence. AAAI: Association for the Advancement of Artificial Intelligence, Technical Tracks, vol. 35, 11525–11535.","chicago":"Grunbacher, Sophie, Ramin Hasani, Mathias Lechner, Jacek Cyranka, Scott A Smolka, and Radu Grosu. “On the Verification of Neural ODEs with Stochastic Guarantees.” In <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>, 35:11525–35. AAAI Press, 2021.","apa":"Grunbacher, S., Hasani, R., Lechner, M., Cyranka, J., Smolka, S. A., &#38; Grosu, R. (2021). On the verification of neural ODEs with stochastic guarantees. In <i>Proceedings of the AAAI Conference on Artificial Intelligence</i> (Vol. 35, pp. 11525–11535). Virtual: AAAI Press.","mla":"Grunbacher, Sophie, et al. “On the Verification of Neural ODEs with Stochastic Guarantees.” <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>, vol. 35, no. 13, AAAI Press, 2021, pp. 11525–35.","ama":"Grunbacher S, Hasani R, Lechner M, Cyranka J, Smolka SA, Grosu R. On the verification of neural ODEs with stochastic guarantees. In: <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>. Vol 35. AAAI Press; 2021:11525-11535.","ieee":"S. Grunbacher, R. Hasani, M. Lechner, J. Cyranka, S. A. Smolka, and R. Grosu, “On the verification of neural ODEs with stochastic guarantees,” in <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>, Virtual, 2021, vol. 35, no. 13, pp. 11525–11535.","short":"S. Grunbacher, R. Hasani, M. Lechner, J. Cyranka, S.A. Smolka, R. Grosu, in:, Proceedings of the AAAI Conference on Artificial Intelligence, AAAI Press, 2021, pp. 11525–11535."},"issue":"13","month":"05","arxiv":1,"file":[{"date_created":"2022-01-26T07:38:08Z","file_size":286906,"date_updated":"2022-01-26T07:38:08Z","creator":"mlechner","file_id":"10680","success":1,"file_name":"17372-Article Text-20866-1-2-20210518.pdf","access_level":"open_access","content_type":"application/pdf","relation":"main_file","checksum":"468d07041e282a1d46ffdae92f709630"}],"department":[{"_id":"GradSch"},{"_id":"ToHe"}],"abstract":[{"lang":"eng","text":"We show that Neural ODEs, an emerging class of timecontinuous neural networks, can be verified by solving a set of global-optimization problems. For this purpose, we introduce Stochastic Lagrangian Reachability (SLR), an\r\nabstraction-based technique for constructing a tight Reachtube (an over-approximation of the set of reachable states\r\nover a given time-horizon), and provide stochastic guarantees in the form of confidence intervals for the Reachtube bounds. SLR inherently avoids the infamous wrapping effect (accumulation of over-approximation errors) by performing local optimization steps to expand safe regions instead of repeatedly forward-propagating them as is done by deterministic reachability methods. To enable fast local optimizations, we introduce a novel forward-mode adjoint sensitivity method to compute gradients without the need for backpropagation. Finally, we establish asymptotic and non-asymptotic convergence rates for SLR."}],"intvolume":"        35","has_accepted_license":"1","file_date_updated":"2022-01-26T07:38:08Z","publication_status":"published","publication_identifier":{"isbn":["978-1-57735-866-4"],"issn":["2159-5399"],"eissn":["2374-3468"]},"title":"On the verification of neural ODEs with stochastic guarantees","oa_version":"Published Version","day":"28","author":[{"first_name":"Sophie","last_name":"Grunbacher","full_name":"Grunbacher, Sophie"},{"full_name":"Hasani, Ramin","last_name":"Hasani","first_name":"Ramin"},{"last_name":"Lechner","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","full_name":"Lechner, Mathias","first_name":"Mathias"},{"last_name":"Cyranka","full_name":"Cyranka, Jacek","first_name":"Jacek"},{"first_name":"Scott A","full_name":"Smolka, Scott A","last_name":"Smolka"},{"last_name":"Grosu","full_name":"Grosu, Radu","first_name":"Radu"}],"date_created":"2022-01-25T15:47:20Z","volume":35,"status":"public","publication":"Proceedings of the AAAI Conference on Artificial Intelligence","project":[{"_id":"25F42A32-B435-11E9-9278-68D0E5697425","grant_number":"Z211","name":"The Wittgenstein Prize","call_identifier":"FWF"}],"acknowledgement":"The authors would like to thank the reviewers for their insightful comments. RH and RG were partially supported by\r\nHorizon-2020 ECSEL Project grant No. 783163 (iDev40). RH was partially supported by Boeing. ML was supported\r\nin part by the Austrian Science Fund (FWF) under grant Z211-N23 (Wittgenstein Award). SG was funded by FWF\r\nproject W1255-N23. JC was partially supported by NAWA Polish Returns grant PPN/PPO/2018/1/00029. SS was supported by NSF awards DCL-2040599, CCF-1918225, and CPS-1446832.\r\n","date_published":"2021-05-28T00:00:00Z","conference":{"start_date":"2021-02-02","end_date":"2021-02-09","name":"AAAI: Association for the Advancement of Artificial Intelligence","location":"Virtual"},"external_id":{"arxiv":["2012.08863"]},"year":"2021","ddc":["000"],"page":"11525-11535","quality_controlled":"1","main_file_link":[{"url":"https://ojs.aaai.org/index.php/AAAI/article/view/17372","open_access":"1"}],"publisher":"AAAI Press","article_processing_charge":"No","alternative_title":["Technical Tracks"],"type":"conference","_id":"10669","date_updated":"2022-05-24T06:33:14Z"},{"publication_status":"published","file_date_updated":"2022-01-26T07:37:24Z","tmp":{"image":"/images/cc_by_nc_nd.png","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","short":"CC BY-NC-ND (3.0)"},"abstract":[{"text":"Imitation learning enables high-fidelity, vision-based learning of policies within rich, photorealistic environments. However, such techniques often rely on traditional discrete-time neural models and face difficulties in generalizing to domain shifts by failing to account for the causal relationships between the agent and the environment. In this paper, we propose a theoretical and experimental framework for learning causal representations using continuous-time neural networks, specifically over their discrete-time counterparts. We evaluate our method in the context of visual-control learning of drones over a series of complex tasks, ranging from short- and long-term navigation, to chasing static and dynamic objects through photorealistic environments. Our results demonstrate that causal continuous-time\r\ndeep models can perform robust navigation tasks, where advanced recurrent models fail. These models learn complex causal control representations directly from raw visual inputs and scale to solve a variety of tasks using imitation learning.","lang":"eng"}],"has_accepted_license":"1","date_created":"2022-01-25T15:47:50Z","oa_version":"Published Version","title":"Causal navigation by continuous-time neural networks","author":[{"last_name":"Vorbach","full_name":"Vorbach, Charles J","first_name":"Charles J"},{"last_name":"Hasani","full_name":"Hasani, Ramin","first_name":"Ramin"},{"first_name":"Alexander","full_name":"Amini, Alexander","last_name":"Amini"},{"first_name":"Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87","full_name":"Lechner, Mathias","last_name":"Lechner"},{"first_name":"Daniela","full_name":"Rus, Daniela","last_name":"Rus"}],"day":"01","user_id":"8b945eb4-e2f2-11eb-945a-df72226e66a9","citation":{"mla":"Vorbach, Charles J., et al. “Causal Navigation by Continuous-Time Neural Networks.” <i>35th Conference on Neural Information Processing Systems</i>, 2021.","apa":"Vorbach, C. J., Hasani, R., Amini, A., Lechner, M., &#38; Rus, D. (2021). Causal navigation by continuous-time neural networks. In <i>35th Conference on Neural Information Processing Systems</i>. Virtual.","chicago":"Vorbach, Charles J, Ramin Hasani, Alexander Amini, Mathias Lechner, and Daniela Rus. “Causal Navigation by Continuous-Time Neural Networks.” In <i>35th Conference on Neural Information Processing Systems</i>, 2021.","ista":"Vorbach CJ, Hasani R, Amini A, Lechner M, Rus D. 2021. Causal navigation by continuous-time neural networks. 35th Conference on Neural Information Processing Systems. NeurIPS: Neural Information Processing Systems,  Advances in Neural Information Processing Systems, .","short":"C.J. Vorbach, R. Hasani, A. Amini, M. Lechner, D. Rus, in:, 35th Conference on Neural Information Processing Systems, 2021.","ieee":"C. J. Vorbach, R. Hasani, A. Amini, M. Lechner, and D. Rus, “Causal navigation by continuous-time neural networks,” in <i>35th Conference on Neural Information Processing Systems</i>, Virtual, 2021.","ama":"Vorbach CJ, Hasani R, Amini A, Lechner M, Rus D. Causal navigation by continuous-time neural networks. In: <i>35th Conference on Neural Information Processing Systems</i>. ; 2021."},"language":[{"iso":"eng"}],"oa":1,"file":[{"date_updated":"2022-01-26T07:37:24Z","creator":"mlechner","file_size":6841228,"date_created":"2022-01-26T07:37:24Z","file_id":"10679","access_level":"open_access","content_type":"application/pdf","success":1,"file_name":"NeurIPS-2021-causal-navigation-by-continuous-time-neural-networks-Paper.pdf","checksum":"be81f0ade174a8c9b2d4fe09590b2021","relation":"main_file"}],"department":[{"_id":"GradSch"},{"_id":"ToHe"}],"arxiv":1,"month":"12","quality_controlled":"1","main_file_link":[{"url":"https://proceedings.neurips.cc/paper/2021/hash/67ba02d73c54f0b83c05507b7fb7267f-Abstract.html","open_access":"1"}],"ddc":["000"],"type":"conference","date_updated":"2022-01-26T14:33:31Z","_id":"10670","article_processing_charge":"No","alternative_title":[" Advances in Neural Information Processing Systems"],"acknowledgement":"C.V., R.H. A.A. and D.R. are partially supported by Boeing and MIT. A.A. is supported by the National Science Foundation (NSF) Graduate Research Fellowship Program. M.L. is supported in part by the Austrian Science Fund (FWF) under grant Z211-N23 (Wittgenstein Award). Research was sponsored by the United States Air Force Research Laboratory and the United States Air Force Artificial Intelligence Accelerator and was accomplished under Cooperative Agreement Number FA8750-19-2-1000. The views and conclusions contained in this document are those of the authors\r\nand should not be interpreted as representing the official policies, either expressed or implied, of the United States Air Force or the U.S. Government. The U.S. Government is authorized to reproduce and distribute reprints for Government purposes notwithstanding any copyright notation herein.\r\n","conference":{"location":"Virtual","name":"NeurIPS: Neural Information Processing Systems","end_date":"2021-12-10","start_date":"2021-12-06"},"date_published":"2021-12-01T00:00:00Z","project":[{"call_identifier":"FWF","name":"The Wittgenstein Prize","grant_number":"Z211","_id":"25F42A32-B435-11E9-9278-68D0E5697425"}],"publication":"35th Conference on Neural Information Processing Systems","status":"public","external_id":{"arxiv":["2106.08314"]},"year":"2021"}]