Research map / Optimization / Nonconvex and smooth optimization
Uniform Convergence with Square-Root Lipschitz Loss
Lijia Zhou, Zhen Dai, Frederic Koehler, Nathan Srebro
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 a nonnegative loss whose square root is Lipschitz instead of requiring the loss to be smooth | Uniform convergence for Gaussian data · Gaussian multi-index models | uniform generalization bound | Improved |
| Analyze phase retrieval with the loss (|ŷ|-y)^2 | Minimal-norm phase-retrieval interpolant · Phase retrieval | population loss consistency | Improved |
| Use a loss that treats positive ReLU labels with squared error and zero labels with σ(ŷ)^2 | Minimal-norm ReLU interpolant · ReLU regression | population loss consistency | Improved |
| Apply the uniform-convergence analysis to the minimal nuclear-norm interpolant | Matrix sensing · Noisy Gaussian matrix sensing | relative squared Frobenius reconstruction error | Improved |
| Use a weighted square loss when the non-Gaussian tail depends on the signal through h | Uniform convergence for dependent non-Gaussian features · Non-Gaussian universality counterexample | weighted population loss | Improved |
| Assume Gaussian universality from covariance matching without accounting for signal-tail dependence | Optimistic-rate generalization analysis · Non-Gaussian features | population squared loss bound | Hurt |
Closest papers by meaning
- Fast Last-Iterate Convergence of SGD in the Smooth Interpolation Regime (NeurIPS 2025 · similarity 0.86)
- Optimizing $(L_0, L_1)$-Smooth Functions by Gradient Methods (ICLR 2025 · similarity 0.86)
- Statistical Guarantees for High-Dimensional Stochastic Gradient Descent (NeurIPS 2025 · similarity 0.86)
- Generalization of noisy SGD in unbounded non-convex settings (ICML 2025 · similarity 0.85)
- Delving into the Convergence of Generalized Smooth Minimax Optimization (ICML 2024 · similarity 0.85)
- Uniform Stability and Generalization Error of GD and SGD on Fixed-Point Parameters (NeurIPS 2026 · 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.
