Research map / Optimization / Nonconvex and smooth optimization
Relative Interior Rule in Block-Coordinate Descent
Check 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 |
|---|---|---|---|
| Choose a relative-interior minimizer when a block has multiple minimizers | Block-coordinate descent · General convex problems | objective value | Helped |
| Choose an endpoint rather than a relative-interior point of the optimal interval (ablation) | Max-sum diffusion · MAP inference dual LP relaxation | local minimum quality | Hurt |
| Choose another point inside the optimal interval instead of its midpoint (ablation) | Max-sum diffusion · MAP inference dual LP relaxation | algorithm performance | No clear effect |
| Use update (29), which minimizes the dual objective subject to the Potts constraints and satisfies the relative interior rule | Potts problem coordinate descent · Toy image segmentation (4 labels, 20 × 20 pixels) | optimality | Helped |
| Use update (37) for single-variable dual optimization | Weighted vertex cover LP relaxation · 41 minimum vertex cover instances from [24] | dual LP optimality | Helped |
| Use update (37) instead of the simplex algorithm | Weighted vertex cover LP relaxation · 41 minimum vertex cover instances from [24] | runtime | Helped |
Closest papers by meaning
- Convergence of Some Convex Message Passing Algorithms to a Fixed Point (ICML 2024 · similarity 0.93)
- Efficient Low Rank Convex Bounds for Pairwise Discrete Graphical Models (ICML 2022 · similarity 0.89)
- Block Acceleration Without Momentum: On Optimal Stepsizes of Block Gradient Descent for Least-Squares (ICML 2024 · similarity 0.86)
- The Sample Complexity of Gradient Descent in Stochastic Convex Optimization (NeurIPS 2024 · similarity 0.86)
- A Study of First-Order Methods with a Deterministic Relative-Error Gradient Oracle (ICML 2024 · similarity 0.86)
- A Universal Transfer Theorem for Convex Optimization Algorithms Using Inexact First-order Oracles (ICML 2024 · 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.