Research map / Learning theory and algorithms / Sample complexity
Robustly Learning a Single Neuron via Sharpness
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 sharpness of the convex surrogate to analyze mini-batch SGD | Single-neuron learning with adversarial label noise · Structured distributions satisfying margin and concentration assumptions | L2^2 loss: O((B^2\alpha^2/(\rho^2 \mu^2)))\opt +\eps | Helped |
| Use the proposed approach rather than the prior convex-surrogate method | Single-neuron learning on logconcave distributions · Logconcave distributions | sample complexity: \Omega(1/\eps^2) versus \polylog(1/\eps) | Helped |
| Require anti-concentration only in an optimal-solution direction and allow inverse-polynomial tails | Distributional assumptions for single-neuron learning · Heavy-tailed and structured discrete distributions | applicability | Helped |
| Replace strong convexity requirements with sharpness | Convex surrogate for robust neuron learning · Heavy-tailed and discrete distributions | applicability | Helped |
| Use refined stochastic-gradient variance scaling with opt + ||w - w*||2^2 | Mini-batch SGD on the surrogate · Surrogate optimization | sample complexity: \polylog(1/\eps) rather than \frac{1}{\epsilon} | Helped |
Closest papers by meaning
- Learning a Single Neuron Robustly to Distributional Shifts and Adversarial Label Noise (NeurIPS 2024 · similarity 0.92)
- Learning a Single Neuron with Bias Using Gradient Descent (NeurIPS 2021 · similarity 0.87)
- Adversarially Robust Learning: A Generic Minimax Optimal Learner and Characterization (NeurIPS 2022 · similarity 0.87)
- Sample and Computationally Efficient Robust Learning of Gaussian Single-Index Models (NeurIPS 2024 · similarity 0.86)
- Robustly Learning Single-Index Models via Alignment Sharpness (ICML 2024 · similarity 0.86)
- Robustly Learning Monotone Single-Index Models (NeurIPS 2025 · similarity 0.85)
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.