Research map / Architectures and efficiency / RNNs and normalization
Beyond the Single Neuron Convex Barrier for Neural Network Certification
Gagandeep Singh, Rupanshu Ganvir, Markus Püschel, Martin Vechev
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 |
|---|---|---|---|
| relax multiple ReLUs jointly (k=2,3) instead of single-neuron triangle (ablation) | kPoly · MNIST ConvSmall first 100 regions | verified regions (#): 20 -> 35 | Improved |
| increase k from 1 to 2 to 3 (ablation) | kPoly · MNIST 9x200 output box volume | output box volume: 4.5272 · 1014 -> 2.9679 · 105 | Improved |
| group neurons by triangle area heuristic instead of random partition (ablation) | kPoly · MNIST ConvSmall first 100 regions | verified regions (#): 34 -> 35 | Improved |
| add k-ReLU refinement to DeepPoly | DeepPoly · MNIST 9x200 | verified regions (#): 259 -> 506 | Improved |
| add k-ReLU refinement to DeepPoly | DeepPoly · MNIST ConvSmall | verified regions (#): 158 -> 347 | Improved |
| add k-ReLU refinement to DeepPoly | DeepPoly · CIFAR10 ResNet | verified regions (#): 243 -> 245 | No clear effect |
| add k-ReLU refinement to DeepPoly | DeepPoly · MNIST ConvBig | verified regions (#): 711 -> 736 | Improved |
| kPoly vs RefineZono | verifier · MNIST ConvBig | verified regions (#): 648 -> 736 | Improved |
| kPoly vs RefineZono | verifier · MNIST 9x200 | verified regions (#): 316 -> 506 | Improved |
Closest papers by meaning
- Expressivity of ReLU-Networks under Convex Relaxations (ICLR 2024 · similarity 0.91)
- 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 0.88)
- SDP-CROWN: Efficient Bound Propagation for Neural Network Verification with Tightness of Semidefinite Programming (ICML 2025 · similarity 0.87)
- Neural Characteristic Activation Analysis and Geometric Parameterization for ReLU Networks (NeurIPS 2024 · similarity 0.87)
- Compelling ReLU Networks to Exhibit Exponentially Many Linear Regions at Initialization and During Training (ICML 2025 · similarity 0.87)
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.
