Research map / Bandits and online learning / Online convex optimization and games
Online Quantile Regression
Yinan Shen, Dong Xia, Wen-Xin Zhou
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 the two-phase statistical stepsize scheme instead of O(1/t) decaying stepsize (ablation) | online sub-gradient descent QR · Simulation d=100, t1.1 noise | relative error | Improved |
| use the two-phase statistical stepsize instead of a constant stepsize (ablation) | online sub-gradient descent QR · Simulation d=100, t1.1 noise | relative error | Improved |
| use quantile regression instead of least squares online under t1.1 noise | online one-sample QR · Simulation d=100, t1.1 noise | convergence: least squares does not converge; QR comparable to Gaussian noise | Improved |
| use online QR instead of online least squares under Gaussian noise | online one-sample QR · Simulation d=100, Gaussian noise | convergence and accuracy | No clear effect |
| use offline QR instead of online learning at n=20000 | quantile regression · Simulation d=100, n=20000, t1.1 noise | error rate | Improved |
| use online QR instead of offline QR at n=50000 | quantile regression · Simulation d=100, n=50000, t1.1 noise | runtime: 520 seconds -> 30 seconds (one-sample), 5 seconds (batch) | Improved |
| vary first-phase geometric decay rate c from 0.5 to 0.05 (ablation) | statistical stepsize scheme · Simulation d=100 | convergence | No clear effect |
| set both second-phase parameters too small (ca=1, cb=0) (ablation) | statistical stepsize scheme · Simulation d=100 | convergence rate | Hurt |
| use online QR instead of OLS | prediction on news popularity · News popularity test set | test error: 9.8752 -> 8.5824 | Improved |
Closest papers by meaning
- Communication-Constrained Distributed Quantile Regression with Optimal Statistical Guarantees (JMLR 2022 · similarity 0.91)
- Non-parametric Quantile Regression via the K-NN Fused Lasso (JMLR 2021 · similarity 0.90)
- Relaxed Quantile Regression: Prediction Intervals for Asymmetric Noise (ICML 2024 · similarity 0.90)
- Distributed High-Dimensional Quantile Regression: Estimation Efficiency and Support Recovery (ICML 2024 · similarity 0.90)
- Fast Nonlinear Vector Quantile Regression (ICLR 2023 · similarity 0.90)
- QuEst: Enhancing Estimates of Quantile-Based Distributional Measures Using Model Predictions (ICML 2025 · 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.
