Research map / Learning theory and algorithms / Kernels and regression theory
Data-Driven Performance Guarantees for Classical and Learned Optimizers
Rajiv Sambharya, Bartolomeo Stellato
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 |
|---|---|---|---|
| use probabilistic data-driven bounds instead of worst-case bounds for classical optimizers | OSQP and SCS fixed-point iterations · Image deblurring and robust Kalman filtering | quantile bound on fixed-point residual | Improved |
| increase number of samples N from 10 to 100 to 1000 (ablation) | sample convergence bound · Image deblurring and robust Kalman filtering | quantile bound tightness: N=10, N=100, N=1000 | Improved |
| optimize PAC-Bayes bound with ALISTA, TiLISTA or GLISTA instead of running ISTA/FISTA | learned ISTA variants · Sparse coding (m=256, n=512, K=10) | NMSE after 10 iterations | Improved |
| use LISTA with its very large number of weights | PAC-Bayes training · Sparse coding (m=256, n=512, K=10) | NMSE bound | Hurt |
| learn warm starts with PAC-Bayes guarantees instead of cold start or nearest-neighbor warm start | L2WS gradient descent · Unconstrained QP (n=20) | fixed-point residual bound | Improved |
| take the better of worst-case guarantee and probabilistic bound | sample convergence bound · fixed-point residual metric | risk bound | Improved |
| add penalty toward target regularizer value B^target in the training objective | PAC-Bayes training problem · learned optimizers | bound quality | Improved |
Closest papers by meaning
- Learning-to-Optimize with PAC-Bayesian Guarantees: Theoretical Considerations and Practical Implementation (JMLR 2025 · similarity 0.93)
- A Generalization Result for Convergence in Learning-to-Optimize (ICML 2025 · similarity 0.91)
- A Closer Look at Learned Optimization: Stability, Robustness, and Inductive Biases (NeurIPS 2022 · similarity 0.88)
- A PAC-Bayesian Analysis of Randomized Learning with Application to Stochastic Gradient Descent (NeurIPS 2017 · similarity 0.88)
- Stochastic Optimization with Arbitrary Recurrent Data Sampling (ICML 2024 · similarity 0.87)
- Kernel-based guarantees for nonlinear parametric models in Bayesian optimization (NeurIPS 2026 · similarity 0.86)
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.
