Research map / Learning theory and algorithms / Kernels and regression theory
Asymptotic normality and confidence intervals for derivatives of 2-layers neural network in the random features model
Yiwei Shen, Pierre C Bellec
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 |
|---|---|---|---|
| Increase network width ratio psi_d toward infinity (overparametrization) | Random-features 2-layer NN CI · Theoretical CI length, ReLU, SNR 2 | squared CI length | Improved |
| Use small ridge penalty (interpolation) instead of larger lambda for wide networks | Random-features 2-layer NN CI · Theoretical CI length, psi_p = 1, SNR 2 | squared CI length | Improved |
| Use activation functions with gamma_1 = 0 instead of ReLU in the high-SNR interpolating regime | Random-features 2-layer NN CI · Infinite width limit, lambda 10^-3, SNR 2 | normalized squared CI length R | Hurt |
| Use ReLU rather than leaky ReLU, Swish, softplus, tanh or sigmoid | Random-features 2-layer NN CI · Infinite width limit, lambda 10^-3, SNR 2 | normalized squared CI length R | Improved |
| Use gamma_1 = 0 activations instead of common activations in the low-SNR interpolating regime | Random-features 2-layer NN CI · Infinite width limit, lambda 10^-3, SNR 0.5 | normalized squared CI length R | Improved |
| Use gamma_1 = 0 activations (Pennington-Worah) to improve performance | Random-features 2-layer NN CI · CI length across psi_p, lambda, rho | normalized squared CI length R | No clear effect |
Closest papers by meaning
- Asymptotics of feature learning in two-layer networks after one gradient-step (ICML 2024 · similarity 0.90)
- Asymptotic Analysis of Two-Layer Neural Networks after One Gradient Step under Gaussian Mixtures Data with Structure (ICLR 2025 · similarity 0.90)
- Feature Averaging: An Implicit Bias of Gradient Descent Leading to Non-Robustness in Neural Networks (ICLR 2025 · similarity 0.87)
- Simplicity Bias and Optimization Threshold in Two-Layer ReLU Networks (ICML 2025 · similarity 0.87)
- Large Stepsize Gradient Descent for Non-Homogeneous Two-Layer Networks: Margin Improvement and Fast Optimization (NeurIPS 2024 · similarity 0.87)
- Analyzing Neural Scaling Laws in Two-Layer Networks with Power-Law Data Spectra (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.
