Research map / Learning theory and algorithms / Combinatorial and search optimization
A Unified View of Piecewise Linear Neural Network Verification
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 BaBSB instead of Reluplex | PL-NN verification · ACAS | success rate: 79.26% -> 98.40% | Helped |
| Add smart branching (ablation) | BaB · ACAS | runtime | Helped |
| Branch on input features rather than ReLU activations (ablation) | BaB · ACAS | runtime | No clear effect |
| Rebuild the convex approximation and recompute bounds at each branch (ablation) | BaB-relusplit · ACAS | runtime | Helped |
| Use Planet-relaxation bounds instead of interval bounds (ablation) | MIP verification · ACAS | runtime | Helped |
| Use the slightly weaker symmetric big-M encoding (ablation) | Planet-feasible · ACAS | runtime | Hurt |
| Optimize the network output and stop when a negative feasible value is found, rather than solve feasibility alone (ablation) | MIPplanet · CollisionDetection | verification accuracy | Helped |
Closest papers by meaning
- VNN: Verification-Friendly Neural Networks with Hard Robustness Guarantees (ICML 2024 · similarity 0.89)
- Verification of Non-Linear Specifications for Neural Networks (ICLR 2019 · similarity 0.89)
- Training Verification-Friendly Neural Networks via Neuron Behavior Consistency (AAAI 2025 · similarity 0.88)
- Certified but Private: Scalable Zero-Knowledge Proofs for the Formal Verification of Neural Networks (NeurIPS 2026 · similarity 0.88)
- Clip-and-Verify: Linear Constraint-Driven Domain Clipping for Accelerating Neural Network Verification (NeurIPS 2025 · similarity 0.87)
- Verifying Properties of Binary Neural Networks Using Sparse Polynomial Optimization (ICLR 2025 · similarity 0.86)
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.