{"file":[{"date_updated":"2022-01-26T07:36:03Z","file_id":"10678","success":1,"date_created":"2022-01-26T07:36:03Z","file_size":4302669,"relation":"main_file","checksum":"0f06995fba06dbcfa7ed965fc66027ff","content_type":"application/pdf","file_name":"16936-Article Text-20430-1-2-20210518 (1).pdf","access_level":"open_access","creator":"mlechner"}],"type":"conference","day":"28","month":"05","article_processing_charge":"No","date_updated":"2022-05-24T06:36:54Z","publication_status":"published","date_published":"2021-05-28T00:00:00Z","issue":"9","department":[{"_id":"GradSch"},{"_id":"ToHe"}],"main_file_link":[{"open_access":"1","url":"https://ojs.aaai.org/index.php/AAAI/article/view/16936"}],"abstract":[{"text":"We introduce a new class of time-continuous recurrent neural network models. Instead of declaring a learning system’s dynamics by implicit nonlinearities, we construct networks of linear first-order dynamical systems modulated via nonlinear interlinked gates. The resulting models represent dynamical systems with varying (i.e., liquid) time-constants coupled to their hidden state, with outputs being computed by numerical differential equation solvers. These neural networks exhibit stable and bounded behavior, yield superior expressivity within the family of neural ordinary differential equations, and give rise to improved performance on time-series prediction tasks. To demonstrate these properties, we first take a theoretical approach to find bounds over their dynamics, and compute their expressive power by the trajectory length measure in a latent trajectory space. We then conduct a series of time-series prediction experiments to manifest the approximation capability of Liquid Time-Constant Networks (LTCs) compared to classical and modern RNNs.","lang":"eng"}],"oa":1,"oa_version":"Published Version","user_id":"2DF688A6-F248-11E8-B48F-1D18A9856A87","date_created":"2022-01-25T15:48:36Z","publisher":"AAAI Press","title":"Liquid time-constant networks","quality_controlled":"1","status":"public","alternative_title":["Technical Tracks"],"page":"7657-7666","author":[{"full_name":"Hasani, Ramin","last_name":"Hasani","first_name":"Ramin"},{"first_name":"Mathias","last_name":"Lechner","full_name":"Lechner, Mathias","id":"3DC22916-F248-11E8-B48F-1D18A9856A87"},{"last_name":"Amini","first_name":"Alexander","full_name":"Amini, Alexander"},{"full_name":"Rus, Daniela","last_name":"Rus","first_name":"Daniela"},{"first_name":"Radu","last_name":"Grosu","full_name":"Grosu, Radu"}],"has_accepted_license":"1","publication":"Proceedings of the AAAI Conference on Artificial Intelligence","external_id":{"arxiv":["2006.04439"]},"publication_identifier":{"isbn":["978-1-57735-866-4"],"eissn":["2374-3468"],"issn":["2159-5399"]},"conference":{"start_date":"2021-02-02","name":"AAAI: Association for the Advancement of Artificial Intelligence","end_date":"2021-02-09","location":"Virtual"},"citation":{"ieee":"R. Hasani, M. Lechner, A. Amini, D. Rus, and R. Grosu, “Liquid time-constant networks,” in <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>, Virtual, 2021, vol. 35, no. 9, pp. 7657–7666.","mla":"Hasani, Ramin, et al. “Liquid Time-Constant Networks.” <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>, vol. 35, no. 9, AAAI Press, 2021, pp. 7657–66.","short":"R. Hasani, M. Lechner, A. Amini, D. Rus, R. Grosu, in:, Proceedings of the AAAI Conference on Artificial Intelligence, AAAI Press, 2021, pp. 7657–7666.","ama":"Hasani R, Lechner M, Amini A, Rus D, Grosu R. Liquid time-constant networks. In: <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>. Vol 35. AAAI Press; 2021:7657-7666.","ista":"Hasani R, Lechner M, Amini A, Rus D, Grosu R. 2021. Liquid time-constant networks. Proceedings of the AAAI Conference on Artificial Intelligence. AAAI: Association for the Advancement of Artificial Intelligence, Technical Tracks, vol. 35, 7657–7666.","apa":"Hasani, R., Lechner, M., Amini, A., Rus, D., &#38; Grosu, R. (2021). Liquid time-constant networks. In <i>Proceedings of the AAAI Conference on Artificial Intelligence</i> (Vol. 35, pp. 7657–7666). Virtual: AAAI Press.","chicago":"Hasani, Ramin, Mathias Lechner, Alexander Amini, Daniela Rus, and Radu Grosu. “Liquid Time-Constant Networks.” In <i>Proceedings of the AAAI Conference on Artificial Intelligence</i>, 35:7657–66. AAAI Press, 2021."},"language":[{"iso":"eng"}],"_id":"10671","intvolume":"        35","volume":35,"year":"2021","acknowledgement":"R.H. and D.R. are partially supported by Boeing. R.H. and R.G. were partially supported by the Horizon-2020 ECSEL\r\nProject grant No. 783163 (iDev40). M.L. was supported in part by the Austrian Science Fund (FWF) under grant Z211-N23 (Wittgenstein Award). A.A. is supported by the National Science Foundation (NSF) Graduate Research Fellowship Program. This research work is partially drawn from the PhD dissertation of R.H.","file_date_updated":"2022-01-26T07:36:03Z","ddc":["000"],"project":[{"call_identifier":"FWF","name":"The Wittgenstein Prize","_id":"25F42A32-B435-11E9-9278-68D0E5697425","grant_number":"Z211"}]}