Research map / Learning theory and algorithms / Matrix and tensor methods
Nonlinear Covariate Balance in Experimental Design
Qing Chen, Peng Zhang
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 |
|---|---|---|---|
| balance second-order moments (quadratic GSW(2)) instead of linear GSW(1) (ablation) | GSW · Quadratic outcomes, d=10, n=100, phi=0.01 | normalized MSE ratio: 5.27 -> 1.00 | Improved |
| use exponential-kernel GSW(inf) instead of linear GSW(1) (ablation) | GSW · Sinusoidal outcomes, d=10, n=100, phi=0.01 | normalized MSE ratio: 6.16 -> 1.00 | Improved |
| use cubic GSW(3) instead of linear GSW(1) (ablation) | GSW · Cubic outcomes, d=10, n=100, phi=0.01 | normalized MSE ratio: 1.09 -> 1.00 | Improved |
| use nonlinear GSW(2) instead of linear GSW(1) on linear outcomes (ablation) | GSW · Linear outcomes, d=10, n=500, phi=0.01 | normalized MSE ratio: 1.00 -> 4.96 | Hurt |
| use nonlinear GSW variants instead of linear GSW(1) for balance of moments 2-5 (ablation) | GSW · d=10, n=100, phi=0.5 | covariate moment imbalance ratio: 3.46 -> 1.00 (moment 2, GSW(2)) | Improved |
| nonlinear GSW variants versus GSW(1) on first-moment balance (ablation) | GSW · d=10, n=100, phi=0.5 | moment 1 imbalance ratio: 1.00 -> 1.62 (GSW(2)) | Hurt |
| use GSW designs instead of Bernoulli/CR/blocking | GSW(1) · d=10, n=100, phi=0.5 | first-moment imbalance ratio: 12.12 -> 1.00 | Improved |
| use nonlinear GSW variants versus Bernoulli on robustness (ablation) | GSW(2) · d=10, n=100, phi=0.5 | robustness ratio: 1.00 -> 1.17 | No clear effect |
| use low-rank (rank-2d) Gram approximation for n > 100 | Nonlinear GSW · simulations n > 100 | runtime / balance | Improved |
| optimal kernel allocation (PSOD) versus GSW(2) runtime | GSW(2) · d=5 | runtime | Improved |
Closest papers by meaning
- Online Balanced Experimental Design (ICML 2022 · similarity 0.85)
- Experimental Design on a Budget for Sparse Linear Models and Applications (ICML 2016 · similarity 0.85)
- Experimental Design for Linear Functionals in Reproducing Kernel Hilbert Spaces (NeurIPS 2022 · similarity 0.84)
- Multivariate Conformal Selection (ICML 2025 · similarity 0.82)
- A Unified Comparative Study with Generalized Conformity Scores for Multi-Output Conformal Regression (ICML 2025 · similarity 0.82)
- FunctionEvolve: Structure-Guided Symbolic Regression with LLMs (NeurIPS 2026 · similarity 0.81)
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.
