Research map / Learning theory and algorithms / Kernels and regression theory
WONDER: Weighted One-shot Distributed Ridge Regression in High Dimensions
Edgar Dobriban, Yue Sheng
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 optimal weights (summing to more than one) instead of naive uniform averaging of local ridge estimators (ablation) | distributed ridge regression · Isotropic covariance, alpha=1, gamma=0.17 | risk: 30-40% decrease in the risk | Improved |
| use WONDER weighted estimator instead of naive average or single-machine local estimator (ablation) | WONDER · Million Song Year Prediction | test MSE | Improved |
| increase number of machines k (data splitting) for WONDER (ablation) | WONDER · Million Song Year Prediction | test MSE | No clear effect |
| use WONDER distributed ridge instead of full-data ridge | ridge regression · Million Song Dataset | computation time: at least 100x in computation time | Improved |
| use smaller regularization than isotropic optimum with AR-1 covariance and k>1 (ablation) | optimally weighted distributed ridge · AR-1 covariance rho=0.9, n=3000, p=500 | distributed risk | Improved |
| increase n and p tenfold at the same aspect ratio (ablation) | theoretical ARE formula · Isotropic simulation, gamma=p/n | agreement with realized relative efficiency: n=1000, p=100 -> n=10000, p=1000 | Improved |
Closest papers by meaning
- One-shot Distributed Ridge Regression in High Dimensions (ICML 2020 · similarity 0.95)
- A Sharper Generalization Bound for Divide-and-Conquer Ridge Regression (AAAI 2019 · similarity 0.88)
- Optimal Rates of Distributed Regression with Imperfect Kernels (JMLR 2021 · similarity 0.88)
- Doubly Distributed Supervised Learning and Inference with High-Dimensional Correlated Outcomes (JMLR 2020 · similarity 0.87)
- Optimal Convergence Rates for Distributed Nystroem Approximation (JMLR 2023 · similarity 0.87)
- An Iterative, Sketching-based Framework for Ridge Regression (ICML 2018 · similarity 0.87)
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.
