Research map / Learning theory and algorithms / Kernels and regression theory
Random Forest Weighted Local Fréchet Regression with Random Objects
Rui Qiu, Zhou Yu, Ruoqing Zhu
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 random forest weighted local linear Frechet regression instead of global Frechet regression on nonlinear responses | GFR · I-2 distributions (p=2, n=100) | MSE: 0.3045 -> 0.0317 | Improved |
| use RFWLLFR instead of RFWLCFR/FRF in low dimension | RFWLCFR/FRF · I-2 distributions (p=2, n=100) | MSE: 0.0410 -> 0.0317 | Improved |
| use RFWLLFR instead of RFWLCFR/FRF in high dimension | RFWLCFR/FRF · I-2 distributions (p=20, n=1000) | MSE: 0.1399 -> 0.1554 | Hurt |
| use forest-weighted methods instead of global regression when the true model is linear | GFR · I-1 distributions (p=2, n=100) | MSE: 0.0014 -> 0.0031 | Hurt |
| use RFWLLFR instead of FRF on SPD matrices (Log-Cholesky) | FRF · II-1 Log-Cholesky (p=5, n=200) | MSE: 0.507 -> 0.397 | Improved |
| use LFR instead of RFWLLFR when p=2 | RFWLLFR · II-1 Log-Cholesky (p=2, n=100) | MSE: 0.128 -> 0.088 | Hurt |
| use RFWLLFR instead of RFWLCFR/FRF in high-dimension SPD setting | RFWLCFR/FRF · II-2 Log-Cholesky (p=20, n=1000) | MSE: 1.251 -> 1.362 | Hurt |
| use RFWLLFR instead of FRF on spherical data | FRF · III-1 spherical (p=10, n=500) | MSE: 0.035012 -> 0.013339 | Improved |
| use RFWLLFR instead of FRF under geodesic distance | RFWLCFR · III-1 spherical (p=10, n=500) | MSE: 0.035018 -> 0.013339 | Improved |
| use shallow trees (depth 3, one random feature) for RFWLLFR in linear setting | RFWLLFR · I-1 distributions | MSE | Improved |
| raise noise from sigma=0.1 to sigma=0.5 in setting I-2 (p=5, n=500) | RFWLLFR vs RFWLCFR/FRF · I-2 distributions (p=5, n=500) | MSE: RFWLLFR 0.0437 -> 0.0946; FRF 0.0515 -> 0.0850 | Mixed |
| use RFWLLFR with depth six vs deeper trees (ablation) | RFWLLFR · I-2 (p=10, n=500) | MSE | Improved |
Closest papers by meaning
- Fréchet Random Forests for Metric Space Valued Regression with Non Euclidean Predictors (JMLR 2024 · similarity 0.91)
- Distributional Random Forests: Heterogeneity Adjustment and Multivariate Distributional Regression (JMLR 2022 · similarity 0.85)
- Optimal Weighted Random Forests (JMLR 2024 · similarity 0.85)
- Fréchet Regression on the Bures-Wasserstein Manifold (NeurIPS 2026 · similarity 0.85)
- Generalized Random Forests Using Fixed-Point Trees (ICML 2025 · similarity 0.84)
- Minimax Rates and Spectral Distillation for Tree Ensembles (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.
