Artykuł w czasopiśmie
Brak miniatury
Licencja
ClosedAccessDostęp zamknięty
 

On the Verification of Neural ODEs with Stochastic Guarantees

Uproszczony widok
cris.lastimport.scopus2024-02-12T20:47:53Z
dc.abstract.enWe show that Neural ODEs, an emerging class of time-continuous neural networks, can be verified by solving a set of global-optimization problems. For this purpose, we introduce Stochastic Lagrangian Reachability (SLR), an abstraction-based technique for constructing a tight Reachtube (an over-approximation of the set of reachable states over 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.
dc.affiliationUniwersytet Warszawski
dc.conference.countryKanada
dc.conference.datefinish2021-02-09
dc.conference.datestart2021-02-02
dc.conference.placeVancouver
dc.conference.seriesNational Conference of the American Association for Artificial Intelligence
dc.conference.seriesNational Conference of the American Association for Artificial Intelligence
dc.conference.seriesshortcutAAAI
dc.conference.shortcutAAAI 2021
dc.conference.weblinkhttps://aaai.org/Conferences/AAAI-21/
dc.contributor.authorGruenbacher, Sophie
dc.contributor.authorHasani, Ramin
dc.contributor.authorLechner, Mathias
dc.contributor.authorCyranka, Jacek
dc.contributor.authorSmolka, Scott A.
dc.contributor.authorGrosu, Radu
dc.date.accessioned2024-01-25T15:49:19Z
dc.date.available2024-01-25T15:49:19Z
dc.date.issued2021
dc.description.financePublikacja bezkosztowa
dc.identifier.doi10.1609/AAAI.V35I13.17372
dc.identifier.urihttps://repozytorium.uw.edu.pl//handle/item/114882
dc.identifier.weblinkhttps://ojs.aaai.org/index.php/AAAI/article/view/17372
dc.languageeng
dc.pbn.affiliationcomputer and information sciences
dc.relation.pages11525-11535
dc.rightsClosedAccess
dc.sciencecloudnosend
dc.subject.enSafety
dc.subject.enRobustness & Trustworthiness
dc.subject.en(Deep) Neural Network Learning Theory
dc.subject.enSampling/Simulation-based Search
dc.titleOn the Verification of Neural ODEs with Stochastic Guarantees
dc.typeJournalArticle
dspace.entity.typePublication