Research map / Optimization / Optimizers
Learning to Warm-Start Fixed-Point Optimization Algorithms
Rajiv Sambharya, Georgina Hall, Brandon Amos, Bartolomeo Stellato
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 |
|---|---|---|---|
| unroll k fixed-point steps in the warm-start training loss (k>0) instead of k=0 (ablation) | learned warm start (fixed-point residual loss) · Unconstrained QP | mean iterations to reach fixed-point residual | Improved |
| learned warm start with k=0 versus cold start (ablation) | learned warm start · Unconstrained QP | iterations to reach residual | No clear effect |
| learned warm start with regression loss and k=5 versus other variants (ablation) | learned warm start · Lasso | fixed-point residual / gain over cold start | Improved |
| use fixed-point residual loss with k=60 instead of regression loss (ablation) | learned warm start · Quadcopter MPC with OSQP | gain over cold start at many evaluation steps | Improved |
| replace cold start with nearest-neighbor warm start | fixed-point solver initialization · parametric optimization benchmarks | iterations to reach residual | No clear effect |
Closest papers by meaning
- Rethinking Warm-Starts with Predictions: Learning Predictions Close to Sets of Optimal Solutions for Faster $\text{L}$-/$\text{L}^\natural$-Convex Function Minimization (ICML 2023 · similarity 0.87)
- Adaptive Partitioning Schemes for Optimistic Optimization (ICML 2025 · similarity 0.84)
- FSNet: Feasibility-Seeking Neural Network for Constrained Optimization with Guarantees (NeurIPS 2025 · similarity 0.84)
- Where Does Warm-Up Come From? Adaptive Scheduling for Norm-Constrained Optimizers (NeurIPS 2026 · similarity 0.83)
- Trust the Direction, Search the Step: Zero-and-First-Order Methods for LLM Fine-Tuning (NeurIPS 2026 · similarity 0.83)
- Solver-Free Decision-Focused Learning for Linear Optimization Problems (NeurIPS 2025 · similarity 0.83)
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.
