Research map / Learning theory and algorithms / Matrix and tensor methods
An Improved Analysis of Alternating Minimization for Structured Multi-Response Regression
Sheng Chen, Arindam Banerjee
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 |
|---|---|---|---|
| Initialize AltMin randomly instead of with OLS, with n >= 40 (ablation) | AltMin with HTP theta-step · Synthetic sparse multi-response regression (p=1000, s=20, m=10, a=0.9) | normalized L2-error | No clear effect |
| Initialize AltMin randomly instead of with OLS, with n < 40 (ablation) | AltMin with HTP theta-step · Synthetic sparse multi-response regression (p=1000, s=20, m=10, a=0.9) | normalized L2-error | Hurt |
| Use AltMin joint noise-covariance estimation instead of ordinary least squares | OLS estimator · Synthetic sparse multi-response regression (m=10, a=0.9) | normalized L2-error | Improved |
| Increase sample size n from 30 to 60 | randomly-initialized AltMin · Synthetic data, error vs iteration | convergence speed | Improved |
| Increase noise correlation a from 0.5 to 0.9 | AltMin · Synthetic data, n=30,40,50,60 | normalized L2-error: a: 0.5 -> 0.9 | Improved |
| Increase noise correlation a so oracle gains more over OLS | oracle vs OLS · Synthetic data, n=60 | e_orc / e_odn ratio | Improved |
| Increase number of responses m from 10 to 18 | AltMin · Synthetic data, a=0.8, n=30,40,60,80 | normalized L2-error: m: 10 -> 18 | No clear effect |
| Increase number of responses m while comparing AltMin with OLS | AltMin · Synthetic data, n=30, a=0.8 | normalized L2-error vs OLS | Hurt |
Closest papers by meaning
- Alternating Estimation for Structured High-Dimensional Multi-Response Models (NeurIPS 2017 · similarity 0.89)
- A convex optimization formulation for multivariate regression (NeurIPS 2020 · similarity 0.87)
- CoLiDE: Concomitant Linear DAG Estimation (ICLR 2024 · similarity 0.86)
- Modular Regression: Improving Linear Models by Incorporating Auxiliary Data (JMLR 2023 · similarity 0.86)
- MixLasso: Generalized Mixed Regression via Convex Atomic-Norm Regularization (NeurIPS 2018 · similarity 0.85)
- New Insights for the Multivariate Square-Root Lasso (JMLR 2022 · 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.
