Research map / Learning theory and algorithms / Prediction and decision losses
Online Robust Regression via SGD on the l1 loss
Scott Pesme, Nicolas Flammarion
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 |
|---|---|---|---|
| run averaged SGD on l1 loss instead of Huber loss with a well-tuned parameter | online robust regression · synthetic oblivious-outlier regression | estimation error convergence | No clear effect |
| run averaged SGD on l1 loss instead of Huber loss with mistuned parameter (5x or 0.2x optimal) | online robust regression · synthetic oblivious-outlier regression | estimation error convergence | Improved |
| use averaged SGD on l1 loss instead of SGD on l2 loss | online robust regression · synthetic oblivious-outlier regression | estimation error | Improved |
| use online SGD on l1 loss instead of offline AdaCRR-GD | online robust regression · synthetic data, ill-conditioned H (kappa=1/d) | convergence speed | Improved |
| ill-condition the feature covariance (kappa=1 to 1/d) | averaged SGD on l1 loss · synthetic data | convergence rate | No clear effect |
| use averaged SGD on l1 loss versus AdaCRR as outlier proportion varies | online robust regression · breakdown point experiment | estimation error | No clear effect |
Closest papers by meaning
- Robust Sparse Regression with Non-Isotropic Designs (NeurIPS 2024 · similarity 0.87)
- List-decodable Linear Regression (NeurIPS 2019 · similarity 0.87)
- Robust Regression Revisited: Acceleration and Improved Estimation Rates (NeurIPS 2021 · similarity 0.86)
- Adversarial Robustness of Nonparametric Regression (NeurIPS 2025 · similarity 0.85)
- Online Linear Regression in Dynamic Environments via Discounting (ICML 2024 · similarity 0.84)
- Information-Computation Tradeoffs for Noiseless Linear Regression with Oblivious Contamination (NeurIPS 2025 · 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.
