Research map / Optimization / Implicit bias and generalization
Optimal Sets and Solution Paths of ReLU 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 |
|---|---|---|---|
| Select an extreme point approximating maximum L2 norm instead of the minimum-L2 solution | Two-layer ReLU optimal set · UCI binary classification datasets | median test accuracy: fertility: 0.66 -> 0.69; horse-colic: 0.75 -> 0.59 | Mixed |
| Choose a solution by minimizing validation MSE over the optimal set instead of minimum L2 norm | Two-layer ReLU optimal set · UCI binary classification datasets | median test accuracy: fertility: 0.66 -> 0.65; pima: 0.68 -> 0.68 | Mixed |
| Choose a solution by minimizing test MSE over the optimal set instead of minimum L2 norm | Two-layer ReLU optimal set · UCI binary classification datasets | median test accuracy: heart-hung.: 0.75 -> 0.85; fertility: 0.66 -> 0.64 | Mixed |
| Use Optimal/LS pruning instead of magnitude-, gradient-, or random-based pruning | Two-layer ReLU networks · UCI binary classification datasets | test accuracy | Helped |
| Use Optimal/LS pruning instead of magnitude-based pruning | Two-layer ReLU networks · hill-valley | test accuracy: approximately 40% | Helped |
| Use Optimal/LS pruning with weight correction after pruning instead of the baselines | Two-layer ReLU networks · Two binary CIFAR-10 tasks | test performance | Helped |
Closest papers by meaning
- Convex Relaxations of ReLU Neural Networks Approximate Global Optima in Polynomial Time (ICML 2024 · similarity 0.93)
- Convex Formulations for Training Two-Layer ReLU Neural Networks (ICLR 2025 · similarity 0.91)
- Compelling ReLU Networks to Exhibit Exponentially Many Linear Regions at Initialization and During Training (ICML 2025 · similarity 0.89)
- Neural Characteristic Activation Analysis and Geometric Parameterization for ReLU Networks (NeurIPS 2024 · similarity 0.89)
- Global Minimizers of $\ell^p$-Regularized Objectives Yield the Sparsest ReLU Neural Networks (NeurIPS 2025 · similarity 0.89)
- Expressivity of ReLU-Networks under Convex Relaxations (ICLR 2024 · similarity 0.89)
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.