Research map / Optimization / Nonconvex and smooth optimization
Alternating Randomized Block Coordinate Descent
Jelena Diakonikolas, Lorenzo Orecchia
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 |
|---|---|---|---|
| add exact minimization over the least smooth block (ablation) | C-BCD and RCDM · BlogFeedback least squares, block sizes 5-40 | optimality gap over epochs | Improved |
| use AR-BCD instead of RCDM and C-BCD | non-accelerated block coordinate descent · BlogFeedback least squares, N/n in {5,10,20,40} | optimality gap over epochs | Improved |
| RCDM versus AR-BCD early iterations | RCDM · BlogFeedback least squares | optimality gap early vs late | Mixed |
| use AAR-BCD and NU ACDM instead of APCG and ABCGD | accelerated block coordinate descent · BlogFeedback least squares | optimality gap over epochs | Improved |
| increase block size N/n from 5 to 40 (ablation) | AAR-BCD · BlogFeedback least squares | improvement over other accelerated methods | Improved |
| use non-uniform sampling proportional to sqrt(Li) instead of uniform | accelerated methods · BlogFeedback least squares | convergence speed | Improved |
Closest papers by meaning
- Block Acceleration Without Momentum: On Optimal Stepsizes of Block Gradient Descent for Least-Squares (ICML 2024 · similarity 0.89)
- Alternating Minimizations Converge to Second-Order Optimal Solutions (ICML 2019 · similarity 0.89)
- Fundamental Benefit of Alternating Updates in Minimax Optimization (ICML 2024 · similarity 0.87)
- Double-Step Alternating Extragradient with Increasing Timescale Separation for Finding Local Minimax Points: Provable Improvements (ICML 2024 · similarity 0.87)
- ADMM for Nonconvex Optimization under Minimal Continuity Assumption (ICLR 2025 · similarity 0.86)
- Block Coordinate Descent for Neural Networks Provably Finds Global Minima (NeurIPS 2025 · 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.
