Research map / Learning theory and algorithms / Combinatorial and search optimization
The Convex Relaxation Barrier, Revisited: Tightened Single-Neuron Relaxations for 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 |
|---|---|---|---|
| add multivariate-input ReLU relaxation inequalities (FastC2V) | DeepPoly · MNIST 6x100 | images verified: 160 -> 279 | Helped |
| add FastC2V inequalities | DeepPoly · MNIST 9x200 | images verified: 259 -> 392 | Helped |
| add FastC2V inequalities | DeepPoly · MNIST ConvBig | images verified: 652 -> 691 | Helped |
| add FastC2V inequalities | DeepPoly · CIFAR-10 ConvSmall | images verified: 359 -> 390 | Helped |
| add cutting planes from the tightened relaxation (OptC2V) | Delta-relaxation LP · MNIST 6x100 | images verified: 201 -> 429 | Helped |
| add OptC2V cuts | Delta-relaxation LP · MNIST ConvBig | images verified: 743 -> 771 | Helped |
| add OptC2V cuts | Delta-relaxation LP · CIFAR-10 ConvSmall | images verified: 373 -> 398 | Helped |
| use FastC2V instead of full Delta-relaxation LP | LP · MNIST 6x100 | images verified: 201 -> 279 | Helped |
| use FastC2V instead of full Delta-relaxation LP | LP · MNIST ConvBig | images verified: 743 -> 691 | Hurt |
| use OptC2V instead of kPoly | kPoly · MNIST 6x100 | images verified: 441 -> 429 | Hurt |
| use OptC2V instead of kPoly | kPoly · MNIST ConvSmall | images verified: 347 -> 436 | Helped |
Closest papers by meaning
- Expressiveness of Multi-Neuron Convex Relaxations in Neural Network Certification (ICLR 2026 · similarity 0.95)
- Expressivity of ReLU-Networks under Convex Relaxations (ICLR 2024 · similarity 0.92)
- SDP-CROWN: Efficient Bound Propagation for Neural Network Verification with Tightness of Semidefinite Programming (ICML 2025 · similarity 0.89)
- Convex Relaxations of ReLU Neural Networks Approximate Global Optima in Polynomial Time (ICML 2024 · similarity 0.89)
- Convex Formulations for Training Two-Layer ReLU Neural Networks (ICLR 2025 · similarity 0.89)
- Training Verification-Friendly Neural Networks via Neuron Behavior Consistency (AAAI 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.