Research map / Learning theory and algorithms / Prediction and decision losses
Towards Constituting Mathematical Structures for Learning to Optimize
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 |
|---|---|---|---|
| fix b1 and b2 biases to zero instead of learning them (ablation) | math-structured L2O on LASSO · LASSO synthetic | optimality gap | Helped |
| fix b_k=1 and do not parameterize b_k (simplified L2O-PA) (ablation) | math-structured L2O on LASSO · LASSO synthetic | optimality gap | No clear effect |
| OOD test on real BSDS500 data: L2O-PA vs L2O-DM and L2O-RNNprop | math-structured L2O (L2O-PA) on LASSO · LASSO train synthetic, test BSDS500 | convergence to tolerance 1e-2 | Helped |
| L2O-PA vs FISTA on logistic regression (in-distribution) | math-structured L2O (L2O-PA) · l1 logistic regression synthetic | iterations to converge: ~100x faster than FISTA | Helped |
| L2O-PA vs FISTA on logistic regression (OOD real data) | math-structured L2O (L2O-PA) · l1 logistic regression real data | iterations to converge: more than 10x faster | Helped |
Closest papers by meaning
- Towards Robust Learning to Optimize with Theoretical Guarantees (CVPR 2024 · similarity 0.90)
- A Generalization Result for Convergence in Learning-to-Optimize (ICML 2025 · similarity 0.87)
- Learning Provably Improves the Convergence of Gradient Descent (NeurIPS 2025 · similarity 0.86)
- Solver-Free Decision-Focused Learning for Linear Optimization Problems (NeurIPS 2025 · similarity 0.86)
- SYMBOL: Generating Flexible Black-Box Optimizers through Symbolic Equation Learning (ICLR 2024 · similarity 0.85)
- ROOT: Rethinking Offline Optimization as Distributional Translation via Probabilistic Bridge (NeurIPS 2025 · similarity 0.85)
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.