Research map / Learning theory and algorithms / Prediction and decision losses
Kernel-based Distributed Learning Beyond Least Squares
Heng Lian, Xu Guo
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 |
|---|---|---|---|
| increase number of machines m at fixed N=1024 (Huber) | one-shot distributed kernel estimator · Nonparametric regression, t2 noise, N=1024 | estimation error: 0.067 -> 0.393 | Hurt |
| increase number of machines m at fixed N=1024 (logistic) | one-shot distributed kernel estimator · Nonparametric logistic model, N=1024 | estimation error: 0.678 -> 6.827 | Hurt |
| increase number of machines m at fixed N=1024 (quantile) | one-shot distributed kernel estimator · Quantile regression tau=0.5, N=1024 | estimation error: 0.089 -> 0.221 | Hurt |
| use Huber loss instead of least squares under heavy-tailed t2 errors | distributed kernel regression · Nonparametric regression, t2 noise, N=1024 | estimation error: 0.173 -> 0.067 | Improved |
| use Huber loss instead of least squares at n=32, m=32 | distributed kernel regression · Nonparametric regression, t2 noise | estimation error: 0.733 -> 0.393 | Improved |
| increase local sample size n at m=1 (Huber) | Huber kernel regression · Nonparametric regression, t2 noise | estimation error: 4.220 -> 0.067 | Improved |
| increase number of machines m at n=32 (Huber) | Huber kernel regression · Nonparametric regression, t2 noise | estimation error: 4.220 -> 0.393 | Improved |
| distribute N=1024 as n=128, m=8 instead of one machine (Huber) | Huber kernel regression · Computation time, Huber loss | seconds: 645.49 -> 93.70 | Improved |
| increase local sample size from n=32 to n=1024 (logistic, real data) | distributed kernel logistic regression · HTRU2 logistic regression (N=4096) | relative error: 0.833 -> 0.145 | Improved |
| increase local sample size from n=32 to n=1024 (quantile tau=0.9) | distributed kernel quantile regression · Air quality quantile regression tau=0.9 (N=4096) | relative error: 0.642 -> 0.150 | Improved |
Closest papers by meaning
- Distributed Learning with Regularized Least Squares (JMLR 2017 · similarity 0.91)
- Optimal Convergence for Distributed Learning with Stochastic Gradient Methods and Spectral Algorithms (JMLR 2020 · similarity 0.90)
- Optimal Distributed Learning with Multi-pass Stochastic Gradient Methods (ICML 2018 · similarity 0.90)
- Distributed Learning of Conditional Quantiles in the Reproducing Kernel Hilbert Space (NeurIPS 2022 · similarity 0.87)
- Approximate Gradient Coding for Distributed Learning with Heterogeneous Stragglers (NeurIPS 2025 · similarity 0.86)
- On Statistical Learning Theory for Distributional Inputs (ICML 2024 · similarity 0.85)
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.
