{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,12,26]],"date-time":"2025-12-26T07:14:52Z","timestamp":1766733292034,"version":"3.41.0"},"reference-count":30,"publisher":"IOP Publishing","issue":"2","license":[{"start":{"date-parts":[[2025,5,30]],"date-time":"2025-05-30T00:00:00Z","timestamp":1748563200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2025,5,30]],"date-time":"2025-05-30T00:00:00Z","timestamp":1748563200000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/iopscience.iop.org\/info\/page\/text-and-data-mining"}],"content-domain":{"domain":["iopscience.iop.org"],"crossmark-restriction":false},"short-container-title":["Mach. Learn.: Sci. Technol."],"published-print":{"date-parts":[[2025,6,30]]},"abstract":"<jats:title>Abstract<\/jats:title>\n               <jats:p>Latent ordinary differential equation (ODE) models provide flexible descriptions of dynamic systems, but they can struggle with extrapolation and predicting complicated non-linear dynamics. The latent ODE approach implicitly relies on encoders to identify unknown system parameters and initial conditions (ICs), whereas the evaluation times are known and directly provided to the ODE solver. This dichotomy can be exploited by encouraging <jats:italic>time-independent<\/jats:italic> latent representations. By replacing the common variational penalty in latent space with an <jats:inline-formula>\n                     <jats:tex-math\/>\n                     <mml:math xmlns:mml=\"http:\/\/www.w3.org\/1998\/Math\/MathML\" overflow=\"scroll\">\n                        <mml:mrow>\n                           <mml:msub>\n                              <mml:mi>\u2113<\/mml:mi>\n                              <mml:mn>2<\/mml:mn>\n                           <\/mml:msub>\n                        <\/mml:mrow>\n                     <\/mml:math>\n                  <\/jats:inline-formula> penalty on the path length of each system, the models learn data representations that can easily be distinguished from those of systems with different configurations. This results in faster training, smaller models, more accurate interpolation and long-time extrapolation compared to the baseline ODE models with a gated recurent unit (GRU), recurrent neural network, and long-short-term-memory encoder\/decoders on tests with damped harmonic oscillator, self-gravitating fluid, and predator-prey systems. We also demonstrate superior results for simulation-based inference of the Lotka\u2013Volterra parameters and ICs by using the latents as data summaries for a conditional normalizing flow. Our change to the training loss is agnostic to the specific recognition network used by the decoder and can therefore easily be adopted by other latent ODE models.<\/jats:p>","DOI":"10.1088\/2632-2153\/addc34","type":"journal-article","created":{"date-parts":[[2025,5,22]],"date-time":"2025-05-22T22:55:37Z","timestamp":1747954537000},"page":"025047","update-policy":"https:\/\/doi.org\/10.1088\/crossmark-policy","source":"Crossref","is-referenced-by-count":1,"title":["Path-minimizing latent ODEs for improved extrapolation and inference"],"prefix":"10.1088","volume":"6","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-5748-5393","authenticated-orcid":true,"given":"Matt L","family":"Sampson","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8873-5065","authenticated-orcid":true,"given":"Peter","family":"Melchior","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"266","published-online":{"date-parts":[[2025,5,30]]},"reference":[{"volume":"vol 31","year":"2018","author":"Chen","key":"mlstaddc34bib1"},{"volume":"vol 32","year":"2019","author":"Rubanova","key":"mlstaddc34bib2"},{"year":"1986","author":"Rumelhart","key":"mlstaddc34bib3"},{"year":"2013","author":"Kingma","key":"mlstaddc34bib4"},{"year":"2020","author":"Lechner","key":"mlstaddc34bib5"},{"key":"mlstaddc34bib6","first-page":"pp 18646","volume":"vol 34","author":"Xia","year":"2021"},{"key":"mlstaddc34bib7","first-page":"pp 9569","article-title":"Segmenting hybrid trajectories using latent odes","author":"Shi","year":"2021"},{"key":"mlstaddc34bib8","doi-asserted-by":"publisher","DOI":"10.1016\/j.amc.2024.128727","volume":"475","author":"Coelho","year":"2024","journal-title":"Appl. Math. Comput."},{"volume":"vol 36","year":"2024","author":"Auzina","key":"mlstaddc34bib9"},{"article-title":"On neural differential equations","year":"2021","author":"Kidger","key":"mlstaddc34bib10"},{"key":"mlstaddc34bib11","first-page":"pp 3154","article-title":"How to train your neural ode: the world of jacobian and kinetic regularization","author":"Finlay","year":"2020"},{"key":"mlstaddc34bib12","first-page":"pp 14831","volume":"vol 33","author":"Ghosh","year":"2020"},{"key":"mlstaddc34bib13","first-page":"pp 4370","volume":"vol 33","author":"Kelly","year":"2020"},{"key":"mlstaddc34bib14","first-page":"pp 18957","article-title":"Minimizing trajectory curvature of ode-based generative models","author":"Lee","year":"2023"},{"key":"mlstaddc34bib15","first-page":"pp 573","article-title":"Invertible residual networks","author":"Behrmann","year":"2019"},{"key":"mlstaddc34bib16","doi-asserted-by":"publisher","first-page":"1735","DOI":"10.1162\/neco.1997.9.8.1735","volume":"9","author":"Schmidhuber","year":"1997","journal-title":"Neural Comput."},{"key":"mlstaddc34bib17","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1016\/0041-5553(64)90137-5","volume":"4","author":"Polyak","year":"1964","journal-title":"Ussr Comput. Math. Math. Phys."},{"year":"2022","author":"Krach","key":"mlstaddc34bib18"},{"year":"2024","author":"Krach","key":"mlstaddc34bib19"},{"key":"mlstaddc34bib20","doi-asserted-by":"publisher","first-page":"2815","DOI":"10.1016\/j.camwa.2010.01.052","volume":"59","author":"Vanani","year":"2010","journal-title":"Comput. Math. Appl."},{"key":"mlstaddc34bib21","doi-asserted-by":"publisher","DOI":"10.1016\/j.ascom.2023.100734","volume":"44","author":"Baty","year":"2023","journal-title":"Astron. Comput."},{"year":"2018","author":"McInnes","key":"mlstaddc34bib22"},{"key":"mlstaddc34bib23","first-page":"pp 1530","article-title":"Variational inference with normalizing flows","author":"Rezende","year":"2015"},{"year":"2019","author":"Winkler","key":"mlstaddc34bib24"},{"year":"2020","author":"Li","key":"mlstaddc34bib25"},{"article-title":"JAX: composable transformations of Python+NumPy programs","year":"2018","author":"Bradbury","key":"mlstaddc34bib26"},{"year":"2021","author":"Kidger","key":"mlstaddc34bib27"},{"article-title":"Flowjax: distributions and normalizing flows in jax","year":"2024","author":"Ward","key":"mlstaddc34bib28"},{"year":"2020","author":"Rackauckas","key":"mlstaddc34bib29"},{"year":"2021","author":"Xia","key":"mlstaddc34bib30"}],"container-title":["Machine Learning: Science and Technology"],"original-title":[],"link":[{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/addc34","content-type":"text\/html","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/addc34\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/addc34","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/addc34\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/addc34\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/addc34\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/addc34\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"similarity-checking"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/addc34\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,5,30]],"date-time":"2025-05-30T10:28:48Z","timestamp":1748600928000},"score":1,"resource":{"primary":{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/addc34"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,5,30]]},"references-count":30,"journal-issue":{"issue":"2","published-online":{"date-parts":[[2025,5,30]]},"published-print":{"date-parts":[[2025,6,30]]}},"URL":"https:\/\/doi.org\/10.1088\/2632-2153\/addc34","relation":{},"ISSN":["2632-2153"],"issn-type":[{"type":"electronic","value":"2632-2153"}],"subject":[],"published":{"date-parts":[[2025,5,30]]},"assertion":[{"value":"Path-minimizing latent ODEs for improved extrapolation and inference","name":"article_title","label":"Article Title"},{"value":"Machine Learning: Science and Technology","name":"journal_title","label":"Journal Title"},{"value":"paper","name":"article_type","label":"Article Type"},{"value":"\u00a9 2025 The Author(s). Published by IOP Publishing Ltd","name":"copyright_information","label":"Copyright Information"},{"value":"2024-10-11","name":"date_received","label":"Date Received","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2025-05-22","name":"date_accepted","label":"Date Accepted","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2025-05-30","name":"date_epub","label":"Online publication date","group":{"name":"publication_dates","label":"Publication dates"}}]}}