Research map / Optimization / Implicit bias and generalization
Theoretical Analysis of Auto Rate-Tuning by Batch Normalization
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 |
|---|---|---|---|
| Allow an arbitrary positive learning rate for scale-invariant parameters while tuning the scale-variant parameters | BatchNorm scale-invariant GD · Smooth nonconvex loss, full-batch gradient descent | stationarity rate: O(T^{-1/2}) | Helped |
| Allow an arbitrary positive initial learning rate for scale-invariant parameters, with α < 1/2 | BatchNorm scale-invariant SGD · Smooth nonconvex loss, stochastic gradient descent | expected squared gradient norm: \tilde{O}\left(\frac{\log T}{\sqrt{T}} \right) | Helped |
| Use α = 1/2 for the scale-invariant learning-rate schedule instead of a slower decay (ablation) | BatchNorm scale-invariant SGD · Smooth nonconvex loss, stochastic gradient descent | expected squared gradient norm: \tilde{O}\left(\frac{\log T}{\sqrt{T}} \right) -> \tilde{O}\left(\frac{(\log T)^{3/2}}{\sqrt{T}}\right) | Hurt |
Closest papers by meaning
- Understanding Batch Normalization (NeurIPS 2018 · similarity 0.91)
- A Quantitative Analysis of the Effect of Batch Normalization on Gradient Descent (ICML 2019 · similarity 0.90)
- Towards Understanding Regularization in Batch Normalization (ICLR 2019 · similarity 0.90)
- Reconciling Modern Deep Learning with Traditional Optimization Analyses: The Intrinsic Learning Rate (NeurIPS 2020 · similarity 0.90)
- Towards Training Without Depth Limits: Batch Normalization Without Gradient Explosion (ICLR 2024 · similarity 0.87)
- Batch Normalization Is Blind to the First and Second Derivatives of the Loss (AAAI 2024 · similarity 0.87)
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.