Research map / Probabilistic and scientific modeling / Dynamical systems
Efficient Verification and Falsification of ReLU Neural Barrier Certificates
Dejin Ren, Yiling Xue, Taoran Wu, Bai Xue
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use necessary-and-sufficient condition over valid linear regions instead of enumerating region intersections | ReLU barrier certificate verification · Arch3 (2-d, hidden 96) | running time (s): 528.36 -> 94.75 | Improved |
| same, on 4-d linear system | ReLU barrier certificate verification · Linear4d (hidden 32) | running time (s): 1110.54 -> 88.12 | Improved |
| same, on 3-d system with two hidden layers of 64 | ReLU barrier certificate verification · Complex (64-64) | verification outcome: baseline crashes (-) vs ours falsified | Improved |
| support falsification in addition to verification | ReLU barrier certificate verification · Arch3 (hidden 128) | validity outcome: baseline unknown (?) vs ours falsified | Improved |
Closest papers by meaning
- SEEV: Synthesis with Efficient Exact Verification for ReLU Neural Barrier Functions (NeurIPS 2024 · similarity 0.88)
- Formal Synthesis of Barrier Certificates Using Fourier Kolmogorov-Arnold Network (AAAI 2025 · similarity 0.87)
- Neural Closure Certificates (AAAI 2024 · similarity 0.87)
- Neural Control and Certificate Repair via Runtime Monitoring (AAAI 2025 · similarity 0.84)
- Provably Safe Neural Network Controllers via Differential Dynamic Logic (NeurIPS 2024 · similarity 0.84)
- Certification from Examples is Hard for Circuits and Transformers under Minimal Overparametrization (NeurIPS 2026 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
