Research map / Label-efficient and robust learning / Recognition and long-tail
Counterexample-Guided Learning of Monotonic Neural Networks
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 counterexample-guided learning (CGL) to baseline NN (ablation) | ReLU NN · Boston Housing (Crime) | MSE: 14.37 -> 11.66 | Helped |
| add counterexample-guided learning (CGL) to baseline NN (ablation) | ReLU NN · Auto-MPG (W,D,HP) | MSE: 9.33 -> 8.63 | Helped |
| add CGL to baseline NN (ablation) | ReLU NN · Heart Disease (Trestbps) | accuracy: 0.85 -> 0.86 | Helped |
| add CGL to baseline NN (ablation) | ReLU NN · Adult | accuracy: 0.84 -> 0.84 | No clear effect |
| add monotonic envelope at prediction time to baseline NN (ablation) | ReLU NN · Auto-MPG (W,D,HP) | MSE: 9.33 -> 9.63 | No clear effect |
| add monotonic envelope at prediction time to baseline NN (ablation) | ReLU NN · Boston Housing (Crime) | MSE: 14.37 -> 14.02 | Helped |
| add CGL on top of envelope (full COMET) (ablation) | NN_b envelope · Boston Housing (Rooms) | MSE: 14.19 -> 11.54 | Helped |
| add CGL on top of envelope (full COMET) (ablation) | NN_b envelope · Heart Disease (Chol.) | accuracy: 0.85 -> 0.87 | Helped |
| COMET versus Min-Max network | COMET · Auto-MPG (Weight) | MSE: 9.91 -> 8.92 | Helped |
| COMET versus DLN | COMET · Boston Housing (Rooms) | MSE: 15.93 -> 11.54 | Helped |
| COMET versus DLN | COMET · Adult (Hours) | accuracy: 0.85 -> 0.84 | Hurt |
| local envelope search versus arbitrary counterexample pair search | envelope construction · Auto MPG | prediction overhead: 48.29 minutes -> a few seconds | Helped |
Closest papers by meaning
- Smooth Min-Max Monotonic Networks (ICML 2024 · similarity 0.88)
- Advancing Constrained Monotonic Neural Networks: Achieving Universal Approximation Beyond Bounded Activations (ICML 2025 · similarity 0.88)
- Enforcing Hard Linear Constraints in Deep Learning Models with Decision Rules (NeurIPS 2025 · similarity 0.85)
- Beyond Interpretability: The Gains of Feature Monosemanticity on Model Robustness (ICLR 2025 · similarity 0.84)
- Simplicity Bias and Optimization Threshold in Two-Layer ReLU Networks (ICML 2025 · similarity 0.84)
- Stable Minima of ReLU Neural Networks Suffer from the Curse of Dimensionality: The Neural Shattering Phenomenon (NeurIPS 2025 · 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.