Research map / Optimization / Optimizers
Relative Fisher Information and Natural Gradient for Learning Large Modular Models
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 |
|---|---|---|---|
| Relative natural gradient descent (layer-wise RFIM preconditioning) instead of SGD/Adam | plain MLP and BN MLP on MNIST · MNIST 784-80-80-80-10 | validation learning curves | Mixed |
| Relative natural gradient descent computational cost relative to SGD and Adam | RNGD · MNIST 784-80-80-80-10 | wall-clock time per epoch: roughly 4 to 10 times more | Hurt |
| Batch normalization combined with RNGD (BN layer added to MLP) | RNGD on MLP · MNIST 784-80-80-80-10 | validation accuracy | Helped |
Closest papers by meaning
- Trade-Offs of Diagonal Fisher Information Matrix Estimators (NeurIPS 2024 · similarity 0.88)
- On the Variance of the Fisher Information for Deep Learning (NeurIPS 2021 · similarity 0.87)
- Fishers for Free? Approximating the Fisher Information Matrix by Recycling the Squared Gradient Accumulator (ICML 2025 · similarity 0.87)
- Fisher information and the geometry of memorization in neural networks (NeurIPS 2026 · similarity 0.86)
- An Improved Empirical Fisher Approximation for Natural Gradient Descent (NeurIPS 2024 · similarity 0.85)
- Bayesian Optimization with Fisher Information Geometry: Gradient Bounds and Trust-Region Methods (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.