Research map / Label-efficient and robust learning / Domain adaptation and generalization
Optimal Parameter-Transfer Learning by Semiparametric Model Averaging
Xiaonan Hu, Xinyu 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 |
|---|---|---|---|
| use semiparametric model averaging over source models instead of target-only least squares | Trans-SMAP · Simulation, correct target model, M=3, n0=150 | averaged out-of-sample MSE: 0.030 -> 0.027 | Improved |
| borrow all source data via pooled least squares instead of averaging weights | LSE-All · Simulation, correct target model, M=3, n0=150 | averaged out-of-sample MSE: 0.027 -> 0.347 | Hurt |
| replace simple averaging weights with the proposed weight criterion | Trans-SimpMA · Simulation, correct target model, M=3, n0=150 | averaged out-of-sample MSE: 0.239 -> 0.027 | Improved |
| use proposed weight criterion instead of BIC weights | Trans-SBIC · Simulation, correct target model, M=3, n0=150 | averaged out-of-sample MSE: 0.188 -> 0.027 | Improved |
| use proposed weight criterion instead of AIC weights | Trans-SAIC · Simulation, correct target model, M=3, n0=150 | averaged out-of-sample MSE: 0.183 -> 0.027 | Improved |
| use semiparametric transfer instead of parametric Trans-Lasso | Trans-SMAP · Simulation, correct target model, M=3, n0=150 | averaged out-of-sample MSE: 0.123 -> 0.027 | Improved |
| use semiparametric transfer instead of parametric Trans-GLM | Trans-SMAP · Simulation, correct target model, M=6, n0=150 | averaged out-of-sample MSE: 0.127 -> 0.026 | Improved |
| misspecified target model, compare to LSE-Tar | Trans-SMAP · Simulation, misspecified target model, M=3, n0=150 | averaged out-of-sample MSE: 0.038 -> 0.034 | Improved |
| increase target sample size n0 from 150 to 500 | Trans-SMAP · Simulation, correct target model, M=3 | averaged out-of-sample MSE: 0.027 -> 0.008 | Improved |
| increase number of candidate models from M=3 to M=6 | Trans-SMAP · Simulation, correct target model, n0=150 | averaged out-of-sample MSE: 0.027 -> 0.026 | No clear effect |
| vary cross-validation folds J (2, 5, 10, leave-one-out) in weight choice (ablation) | Trans-SMAP · Beijing housing rental data | averaged out-of-sample MSE | No clear effect |
| use Trans-SMAP instead of Trans-SimpMA, Trans-SBIC, Trans-SAIC | Trans-SMAP · Beijing housing rental data (MSPE) | averaged out-of-sample MSE | Improved |
Closest papers by meaning
- Enhancing Transfer Learning with Flexible Nonparametric Posterior Sampling (ICLR 2024 · similarity 0.86)
- Wasserstein Transfer Learning (NeurIPS 2025 · similarity 0.84)
- Transfer Learning of Linear Regression with Multiple Pretrained Models: Benefiting from More Pretrained Models via Overparameterization Debiasing (NeurIPS 2026 · similarity 0.84)
- A High-Dimensional Statistical Method for Optimizing Transfer Quantities in Multi-Source Transfer Learning (NeurIPS 2025 · similarity 0.84)
- H-ensemble: An Information Theoretic Approach to Reliable Few-Shot Multi-Source-Free Transfer (AAAI 2024 · similarity 0.84)
- On Hypothesis Transfer Learning of Functional Linear Models (ICML 2024 · similarity 0.84)
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.
