Research map / Learning theory and algorithms / Prediction and decision losses
Parallel and Distributed Block-Coordinate Frank-Wolfe Algorithms
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 |
|---|---|---|---|
| Use asynchronous updates instead of waiting for the slowest worker | BCFW on OCR structural SVM · OCR structural SVM, straggler simulation | average time per effective datapass | Helped |
| Use asynchronous updates with heterogeneous worker speeds | AP-BCFW · OCR structural SVM, heterogeneous-worker simulation | slowdown relative to no-straggler case: 1.4 | Helped |
| Increase mini-batch size up to approximately 50 (ablation) | AP-BCFW · OCR sequence labeling | speedup over BCFW | Helped |
| Increase mini-batch size beyond approximately 50 (ablation) | AP-BCFW · OCR sequence labeling | speedup over BCFW | Mixed |
| Increase mini-batch size to τ ≤ 55 (ablation) | AP-BCFW · Group Fused Lasso, synthetic | speedup over BCFW | Helped |
| Increase mini-batch size beyond τ=55 (ablation) | AP-BCFW · Group Fused Lasso, synthetic | speedup over BCFW | Mixed |
| Increase mini-batch size up to τ=3T (ablation) | AP-BCFW · OCR structural SVM, T=8 | convergence per second | Helped |
| Increase mini-batch size to τ=5T (ablation) | AP-BCFW · OCR structural SVM, T=8 | convergence per second | Hurt |
| Use asynchronous updates rather than synchronous updates that wait for all workers | BCFW on OCR structural SVM · OCR structural SVM, one straggler | average time per effective datapass | Helped |
| Accept delayed updates, dropping those delayed more than k/2 | BCFW on Group Fused Lasso · Group Fused Lasso, Poisson and Pareto delays | iterations to convergence: fewer than twice | Hurt |
| Use longer subproblems (ablation) | AP-BCFW · OCR structural SVM, shared-memory parallelism | speedup with parallel workers | Helped |
Closest papers by meaning
- Beyond Short Steps in Frank-Wolfe Algorithms (ICLR 2026 · similarity 0.89)
- Sarah Frank-Wolfe: Methods for Constrained Optimization with Best Rates and Practical Features (ICML 2024 · similarity 0.88)
- Stochastic Frank-Wolfe for Constrained Finite-Sum Minimization (ICML 2020 · similarity 0.88)
- Accelerated Methods with Compressed Communications for Distributed Optimization Problems Under Data Similarity (AAAI 2025 · similarity 0.85)
- Revisiting Frank-Wolfe for Structured Nonconvex Optimization (NeurIPS 2025 · similarity 0.85)
- Efficient Quadratic Corrections for Frank-Wolfe Algorithms (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.