Research map / Learning theory and algorithms / Sample complexity
Two-Stage Approach to Multivariate Linear Regression with Sparsely Mismatched Data
Martin Slawski, Emanuel Ben-David, Ping Li
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 re-fitting after removing estimated mismatches (proposed+ vs proposed) on case study (ablation) | proposed · Beijing air quality case study | R2: 0.70 -> 0.712 | Improved |
| add re-fitting after removing estimated mismatches (proposed+ vs proposed) on case study (ablation) | proposed · Beijing air quality case study (B* RMSE with intercepts) | B* RMSE: 318.1 -> 295.81 | Improved |
| apply the proposed estimator instead of naive least squares | naive least squares · Beijing air quality case study | B* RMSE: 431.4 -> 318.1 | Improved |
| add second-stage matching correction and refit after proposed first stage (ablation) | proposed · Beijing air quality case study | R2: 0.70 -> 0.715 | Improved |
| apply the proposed estimator instead of naive least squares on US cities data | naive least squares · US cities temperatures (n=46, k=10) | R2: 0.4 -> 0.62 | Improved |
| use the proposed estimator instead of EM in case study | EM · Beijing air quality case study | R2: 0.625 -> 0.70 | Improved |
| add doubly stochastic (DS) convex refinements instead of the group-lasso proposed method | proposed · synthetic data (n in {200,500}, d/n=0.03) | estimation error of B* | No clear effect |
| add refitting step (proposed+) to proposed on synthetic data (ablation) | proposed · synthetic data | standardized estimation error | Improved |
| increase the number of data points n for EM | EM · synthetic data | estimation error of B* | Hurt |
| raise the decay rate of singular values of B* (smaller stable rank) for EM versus proposed+ | EM · synthetic data | estimation error of B* | Hurt |
| use EM instead of naive least squares on real data ftp | naive least squares · real data sarcos/ftp/scm | permutation recovery error | Hurt |
| vary multiplier M in lambda over 0.25, 0.5, 1, 2 (ablation) | proposed · real data sarcos/ftp/scm | permutation recovery error: M in {0.25, 0.5, 1, 2} | No clear effect |
Closest papers by meaning
- Regression with Label Permutation in Generalized Linear Model (ICML 2023 · similarity 0.88)
- Linear Regression With Unmatched Data: A Deconvolution Perspective (JMLR 2024 · similarity 0.86)
- Linear Regression Under Misalignment: Algorithms and Theoretical Results (NeurIPS 2026 · similarity 0.86)
- Optimal Estimator for Unlabeled Linear Regression (ICML 2020 · similarity 0.86)
- Permuted and Unlinked Monotone Regression in R^d: an approach based on mixture modeling and optimal transport (JMLR 2024 · similarity 0.84)
- Regression Learning with Limited Observations of Multivariate Outcomes and Features (ICML 2024 · similarity 0.83)
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.
