Research map / Learning theory and algorithms / Prediction and decision losses
The Robustness of Estimator Composition
Pingfan Tang, Jeff M Phillips
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 |
|---|---|---|---|
| 10% outliers in 10% of streams, E1 10% then E2 10% composite | percentile composite · Simulation 2 | command center output: -1.3327 -> -100.6573 | Improved |
| E1 10% / E2 90% composite needs ~10% outliers to be moved | percentile composite · Simulation 2 | command center output: -1.2169 -> 100.0553 | Improved |
| median-of-medians with 26.01% outliers | median composite · Simulation 2 | command center output: -0.0085 -> 100.1062 | Improved |
| 1.21% outliers on E1 10% / E2 median-type composite (median/median) | median composite · Simulation 2 | command center output: -0.0085 -> 0.0091 | No clear effect |
| 1.21% outliers on E1 10% / E2 90% composite | percentile composite · Simulation 2 | command center output: -1.2169 -> -1.2033 | No clear effect |
| raising outliers from 1.21% to 10.01% for E1 90%/E2 90% | percentile composite · Simulation 2 | command center output: 100.5666 -> 108.6957 | Mixed |
| increase n and k from 5x8 to 1000x200 in L1-median modification | gradient descent algorithm · Simulation 1 | error in x0': 10.7631 -> 10.7025 | Mixed |
| n=1000, k=200 case vs smaller cases | gradient descent algorithm · Simulation 1 | error in x0': 18.6351 -> 18.7886 | Hurt |
| tighter terminating condition (gradient norm < 0.00001) | gradient descent algorithm · Simulation 1 | accuracy of target: 0.00001 | Improved |
| change only 12 of 40 points | modification algorithm · Simulation 1 n=5 k=8 | distance to target: 0.9961, 1.0126 -> 0.9961, 1.0126 | Improved |
Closest papers by meaning
- Robustness Auditing for Linear Regression: To Singularity and Beyond (ICLR 2025 · similarity 0.86)
- All-Purpose Mean Estimation over R: Optimal Sub-Gaussianity with Outlier Robustness and Low Moments Performance (ICML 2025 · similarity 0.85)
- Achievable distributional robustness when the robust risk is only partially identified (NeurIPS 2024 · similarity 0.85)
- Robust Sparse Regression with Non-Isotropic Designs (NeurIPS 2024 · similarity 0.83)
- Adversarial Robustness of Nonparametric Regression (NeurIPS 2025 · similarity 0.83)
- Density Ratio Estimation with Doubly Strong Robustness (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.
