Research map / Architectures and efficiency / RNNs and normalization
The future is log-Gaussian: ResNets and their infinite-depth-and-width limit at initialization
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 Balanced ResNets instead of Vanilla ResNets | ReLU ResNet at initialization · d/n=1 | variance contribution of interlayer correlation: e^{5.5} \approx 250 -> e^{2.5} \approx 10 | Helped |
| log-Gaussian limit instead of Gaussian infinite-width limit | finite ResNet output at initialization · Monte Carlo simulations | agreement with finite-network mean and variance | Helped |
Closest papers by meaning
- Neural signature kernels as infinite-width-depth-limits of controlled ResNets (ICML 2023 · similarity 0.87)
- Infinite-Width Limit of a Single Attention Layer: Analysis via Tensor Programs (NeurIPS 2025 · similarity 0.87)
- Characterizing ResNet's Universal Approximation Capability (ICML 2024 · similarity 0.86)
- Generalization of Scaled Deep ResNets in the Mean-Field Regime (ICLR 2024 · similarity 0.86)
- Feature Learning Dynamics in Infinite-Depth Neural Networks (NeurIPS 2026 · similarity 0.85)
- Quantitative convergence of trained neural networks to Gaussian processes (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.