Research map / Optimization / Optimizers
Conic Descent and its Application to Memory-efficient Optimization over Positive Semidefinite Matrices
John C. Duchi, Oliver Hinder, Andrew Naber, Yinyu Ye
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 conic descent instead of conditional gradient (CG) with overestimate R = 1^T b of the trace bound | phase retrieval on grayscale CIFAR-10 · CIFAR-10 images, SNR 20 dB | matrix multiplications to reach CG objective after 500 iterations | Improved |
| use tighter trace overestimate R = 2/k 1^T b for CG (ablation) | conditional gradient baseline · phase retrieval, CIFAR-10 | iterations to desired objective | Hurt |
| add greedy step every 100 iterations to CD (ablation) | conic descent · PSD matrix completion | objective f(Xk)-p* | Improved |
| greedy step rank r = 2 or 3 vs r = 4 or 5 compared with pure Burer-Monteiro (PBM) | CD without greedy step · PSD matrix completion | objective at equal matrix multiplications | Mixed |
Closest papers by meaning
- No Spurious Local Minima in Nonconvex Low Rank Problems: A Unified Geometric Analysis (ICML 2017 · similarity 0.87)
- A Conditional Gradient Framework for Composite Convex Minimization with Applications to Semidefinite Programming (ICML 2018 · similarity 0.87)
- Rank-1 Matrix Completion with Gradient Descent and Small Random Initialization (NeurIPS 2023 · similarity 0.87)
- Memory-Constrained Algorithms for Convex Optimization (NeurIPS 2023 · similarity 0.87)
- Symmetric Matrix Completion with ReLU Sampling (ICML 2024 · similarity 0.86)
- Spectral Preconditioning for Gradient Methods on Graded Non-convex Functions (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.
