Lune

NeurIPS2022Top-tier venue

Neural Abstractions

Alessandro Abate, Alec Edwards, Mirco Giacobbe

Abstract

We present a novel method for the safety verification of nonlinear dynamical models that uses neural networks to represent abstractions of their dynamics. Neural networks have extensively been used before as approximators; in this work, we make a step further and use them for the first time as abstractions. For a given dynamical model, our method synthesises a neural network that overapproximates its dynamics by ensuring an arbitrarily tight, formally certified bound on the approximation error. For this purpose, we employ a counterexample-guided inductive synthesis procedure. We show that this produces a neural ODE with non-deterministic disturbances that constitutes a formal abstraction of the concrete model under analysis. This guarantees a fundamental property: if the abstract model is safe, i.e., free from any initialised trajectory that reaches an undesirable state, then the concrete model is also safe. By using neural ODEs with ReLU activation functions as abstractions, we cast the safety verification problem for nonlinear dynamical models into that of hybrid automata with affine dynamics, which we verify using SpaceEx. We demonstrate that our approach performs comparably to the mature tool Flow* on existing benchmark nonlinear models. We additionally demonstrate and that it is effective on models that do not exhibit local Lipschitz continuity, which are out of reach to the existing technologies.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

2 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • Provably Safe Neural Network Controllers via Differential Dynamic Logic cites Verification of Neural-Network Control Systems by Integrating Taylor Models and Zonotopes
  • Provably Safe Neural Network Controllers via Differential Dynamic Logic cites Improved Geometric Path Enumeration for Verifying ReLU Neural Networks
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers2

  1. Provably Safe Neural Network Controllers via Differential Dynamic LogicSamuel Teuber, Stefan Mitsch, André PlatzerNeurIPS 2024 · 22 citations
  2. Bisimulation LearningAlessandro Abate, Mirco Giacobbe, Yannik SchnitzerCAV 2024 · 6 citations

Builds on9

  1. Neural Lyapunov Control of Unknown Nonlinear Systems with Stability GuaranteesRuikun Zhou, Thanin Quartz, Hans De Sterck, Jun LiuNeurIPS 2022 · 109 citations
  2. Improved Geometric Path Enumeration for Verifying ReLU Neural NetworksStanley Bak, Hoang-Dung Tran, Kerianne Hobbs, Taylor T. JohnsonCAV 2020 · 88 citations
  3. Verification of Neural-Network Control Systems by Integrating Taylor Models and ZonotopesChristian Schilling, Marcelo Forets, Sebastián GuadalupeAAAI 2022 · 48 citations
  4. Stability Verification in Stochastic Control Systems via Neural Network SupermartingalesMathias Lechner, Dorde Zikelic, Krishnendu Chatterjee, Thomas A. HenzingerAAAI 2022 · 45 citations
  5. On the Verification of Neural ODEs with Stochastic GuaranteesSophie Gruenbacher, Ramin M. Hasani, Mathias Lechner, Jacek Cyranka et al.AAAI 2021 · 37 citations

Related papers

  1. Let a Neural Network be Your InvariantMirco Giacobbe, Daniel Kroening, Abhinandan Pal, Michael TautschnigNeurIPS 2025 · 6 citations
  2. An Iterative Scheme of Safe Reinforcement Learning for Nonlinear Systems via Barrier Certificate GenerationZhengfeng Yang, Yidan Zhang, Wang Lin, Xia Zeng et al.CAV 2021 · 15 citations
  3. SEEV: Synthesis with Efficient Exact Verification for ReLU Neural Barrier FunctionsHongchao Zhang, Zhizhen Qin, Sicun Gao, Andrew ClarkNeurIPS 2024 · 17 citations
  4. An Abstraction-Based Framework for Neural Network VerificationYizhak Yisrael Elboher, Justin Gottschlich, Guy KatzCAV 2020 · 97 citations
  5. Synthesizing Barrier Certificates of Neural Network Controlled Continuous Systems via ApproximationsMeng Sha, Xin Chen, Yuzhe Ji, Qingye Zhao et al.DAC 2021 · 14 citations
Dusk over the sea between two cliffs drawn in fine vertical lines