Research map / Optimization / Nonconvex and smooth optimization
Dykstra's Algorithm, ADMM, and Coordinate Descent: Connections, Insights, and Extensions
Ryan J Tibshirani
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 |
|---|---|---|---|
| Apply block coordinate descent to the regularized regression problem with separable seminorm penalties | Regularized regression · Dual best-approximation problem | algorithmic equivalence | Improved |
| Use ADMM on the best-approximation problem with two sets, one a linear subspace, and the stated initialization and parameter | Dykstra's algorithm · Two-set best-approximation problem | algorithmic equivalence | Improved |
| Transfer polyhedral Dykstra convergence analysis to cyclic coordinate descent | Lasso coordinate descent · Lasso | asymptotic convergence rate | Improved |
| Use parallel-Dykstra coordinate descent instead of serial coordinate descent | Regularized regression with separable seminorm penalties · Regularized regression | convergence to a solution | Improved |
| Use parallel-ADMM coordinate descent instead of serial coordinate descent | Regularized regression with separable seminorm penalties · Regularized regression | convergence to a solution | Improved |
| Increase the ADMM parameter in parallel-ADMM coordinate descent | Parallel-ADMM-CD · Lasso simulations | suboptimality | Improved |
| Use parallel coordinate updates rather than serial updates under the paper's hypothetical 10%-efficient parallel implementation | Parallel-Dykstra-CD and parallel-ADMM-CD · Lasso simulations | suboptimality versus parallelized iteration cost | Improved |
| Use parallel-Dykstra-CD rather than serial coordinate descent | Lasso · Lasso | asymptotic convergence rate | Hurt |
| Extend the Dykstra-coordinate descent correspondence to Bregman projection and nonquadratic loss | Regularized estimation · Nonquadratic-loss problems | algorithmic equivalence | Improved |
| Use warm starts across a decreasing sequence of lasso tuning parameters | Lasso coordinate descent · Lasso | iteration count to enter the linear-convergence phase | Improved |
Closest papers by meaning
- Discerning the Linear Convergence of ADMM for Structured Convex Optimization through the Lens of Variational Analysis (JMLR 2020 · similarity 0.89)
- Smooth Primal-Dual Coordinate Descent Algorithms for Nonsmooth Convex Optimization (NeurIPS 2017 · similarity 0.88)
- Coordinate Descent Methods for DC Minimization: Optimality Conditions and Global Convergence (AAAI 2023 · similarity 0.88)
- Relative Interior Rule in Block-Coordinate Descent (CVPR 2020 · similarity 0.88)
- Directional Smoothness and Gradient Methods: Convergence and Adaptivity (NeurIPS 2024 · similarity 0.86)
- Near-Optimal Solutions of Constrained Learning Problems (ICLR 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.
