Research map / Optimization / Optimizers
Sever: A Robust Meta-Algorithm for Stochastic Optimization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 top-singular-vector projections of centered per-example gradients to score outliers | Sever with SVM on Enron spam · Enron spam, 1% outliers | test error: 7.34% vs 13.43%-20.48% | Helped |
| Use top-singular-vector projections of centered per-example gradients to score outliers | Sever with ridge regression · Drug discovery, 10% corruptions | test error (mean-squared error): 1.42 vs 1.51-2.33 | Helped |
| Use 3 filtering iterations instead of 2 (ablation) | Sever with SVM · Enron spam, 6 difficult attacks | test error: at most 7.4% | Helped |
| Center gradients separately and remove points separately for each class (ablation) | Sever with SVM · Classification datasets | test error | Helped |
| Center the data points (ablation) | Ridge regression · Drug discovery | test error | Helped |
| Scale each coordinate to have variance 1 (ablation) | Ridge regression · Synthetic Gaussian dataset | test error | Helped |
| Use 2 filtering iterations instead of 3 (ablation) | Sever with SVM · Enron spam, 6 difficult attacks | test error | Hurt |
Closest papers by meaning
- From Inexact Gradients to Byzantine Robustness: Acceleration and Optimization under Similarity (NeurIPS 2026 · similarity 0.84)
- Group Distributionally Robust Dataset Distillation with Risk Minimization (ICLR 2025 · similarity 0.83)
- Robust Model-Based Optimization for Challenging Fitness Landscapes (ICLR 2024 · similarity 0.83)
- Robust Sparse Regression with Non-Isotropic Designs (NeurIPS 2024 · similarity 0.83)
- Online Linear Classification with Massart Noise (ICML 2025 · similarity 0.83)
- Boosting Test Performance with Importance Sampling--a Subpopulation Perspective (AAAI 2025 · similarity 0.83)
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.