Research map / Learning theory and algorithms / Sample complexity
Statistical analysis for a penalized EM algorithm in high-dimensional mixture linear regression model
Ning Wang, Xin Zhang, Qing Mai
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 |
|---|---|---|---|
| add post-selection refit (PSEM) after penalized EM (ablation) | PEM · M1 (p=1000) | beta estimation error: 1.26 -> 0.76 | Improved |
| add post-selection refit (PSEM) after penalized EM (ablation) | PEM · M4 (p=400, 3 mixtures) | beta estimation error: 2.43 -> 1.80 | Improved |
| run penalized EM iterations from the tensor-power initialization (ablation) | Initial · M1 (p=400) | beta estimation error: 5.14 -> 1.04 | Improved |
| run penalized EM iterations from the tensor-power initialization (ablation) | Initial · M1 (p=1000) | variable selection TPR: 76.4 -> 100 | Improved |
| use group-lasso penalized EM instead of HDEM | HDEM · M4 (p=400, 3 mixtures) | beta estimation error: 8.87 -> 2.43 | Improved |
| use group-lasso penalized EM instead of HDEM | HDEM · M2 (p=1000) | mixture estimation error: 29.65 -> 23.28 | Improved |
| use penalized EM with variable selection instead of GLLiM without selection | GLLiM · M1 (p=400) | beta estimation error: 9.63 -> 1.04 | Improved |
| use penalized EM instead of HDEM for false positive control | HDEM · M4 (p=400, 3 mixtures) | false positive rate: 12.6 -> 2.7 | Improved |
| input variance sigma^2 of 2 instead of true value 1 in algorithm (ablation) | PEM · delta=1, p=400 | beta estimation error: 1.03 -> 1.23 | Hurt |
| input variance sigma^2 of 0.5 instead of true value 1 in algorithm (ablation) | PEM · delta=0.75, p=400 | beta estimation error: 1.60 -> 1.50 | Improved |
| input variance sigma^2 of 2 instead of true value 1 in algorithm (ablation) | PEM · delta=0.75, p=400 | beta estimation error: 1.60 -> 3.28 | Hurt |
| model multiple responses jointly instead of separately | multivariate mixture linear regression · simulation (q=2, n=400, p=100) | coefficient and mixture estimation error | Improved |
Closest papers by meaning
- Unveiling the Cycloid Trajectory of EM Iterations in Mixed Linear Regression (ICML 2024 · similarity 0.87)
- Big Learning Expectation Maximization (AAAI 2024 · similarity 0.87)
- Agnostic Learning of Mixed Linear Regressions with EM and AM Algorithms (ICML 2024 · similarity 0.86)
- Heterogeneous Sufficient Dimension Reduction and Subspace Clustering (ICML 2025 · similarity 0.86)
- Toward Global Convergence of Gradient EM for Over-Paramterized Gaussian Mixture Models (NeurIPS 2024 · similarity 0.86)
- Is $\sqrt{d}$ Separation Necessary for Gradient EM to Learn Gaussian Mixtures in High Dimensions? (NeurIPS 2026 · 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.
