Research map / Learning theory and algorithms / Sample complexity
Learning Distributions Generated by Single-Layer ReLU Networks in the Presence of Arbitrary Outliers
Saikiran Bulusu, Geethu Joseph, M. Cenk Gursoy, Pramod Varshney
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 |
|---|---|---|---|
| Add median filter to GD (ablation) | GD estimator · Simulation p=0.95, d=5, N=20000 | runtime: 17.73 s -> 34.44 s | Hurt |
| Use trimmed-mean filter instead of median filter in GD | GD estimator · Simulation | runtime: 34.44 s -> 60.78 s | Hurt |
| Use SGD instead of GD (median filter) | Median-filter estimator · Simulation | runtime: 34.44 s -> 2.11 s | Improved |
| Use SGD instead of GD (trimmed mean) | Trimmed-mean estimator · Simulation | runtime: 60.78 s -> 3.62 s | Improved |
| Add filter (median or trimmed mean) to GD or SGD | Estimators with outliers · Simulation, varying p | estimation error | Improved |
| Median-based instead of trimmed-mean-based filter | GD · Simulation | estimation error | Improved |
| Use SGD instead of GD | Filtered estimators · Simulation, Figure 2 | estimation error | Hurt |
| Increase uncorrupted probability p towards 1 | All non-oracle schemes · Simulation | estimation error | Improved |
| Increase number of samples N without a filter | GD/SGD without filter · Simulation | estimation error | No clear effect |
| Increase dimension d up to 500 | Proposed schemes and oracle · Simulation | estimation error | Hurt |
Closest papers by meaning
- Learning Distributions Generated by One-Layer ReLU Networks (NeurIPS 2019 · similarity 0.91)
- Training Two-Layer ReLU Networks with Gradient Descent is Inconsistent (JMLR 2022 · similarity 0.88)
- Benign overfitting in leaky ReLU networks with moderate input dimension (NeurIPS 2024 · similarity 0.86)
- Simplicity Bias and Optimization Threshold in Two-Layer ReLU Networks (ICML 2025 · similarity 0.86)
- Outliers with Opposing Signals Have an Outsized Effect on Neural Network Optimization (ICLR 2024 · similarity 0.86)
- Feature Averaging: An Implicit Bias of Gradient Descent Leading to Non-Robustness in Neural Networks (ICLR 2025 · similarity 0.86)
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.
