Research map / Optimization / Optimizers
Large-Scale Distributed Second-Order Optimization Using Kronecker-Factored Approximate Curvature for Deep Convolutional Neural Networks
Check 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 K-FAC rather than first-order SGD for large-mini-batch training | ResNet-50 · ImageNet, batch sizes 4,096–131,072 | Top-1 accuracy | Helped |
| Refresh Fisher information less frequently after 500 iterations | Distributed K-FAC · ImageNet, batch size 32,768 | Training time | Helped |
| Approximate batch-normalization Fisher matrices as diagonal (ablation) | K-FAC · ImageNet, batch size 32,768 | Top-1 accuracy | No clear effect |
| Approximate batch-normalization Fisher matrices as diagonal (ablation) | K-FAC · ResNet-50 Fisher matrices | Fisher matrix memory: 1017MiB to 587MiB | Helped |
| Use half precision for Kronecker-factor computation | Distributed K-FAC · Kronecker-factor construction | Computation speed: more than doubles | Helped |
| Communicate only the upper triangle of symmetric Kronecker factors | Distributed K-FAC · Kronecker-factor communication | Communication volume: N(N + 1)/2 | Helped |
Closest papers by meaning
- Distributed Second-Order Optimization using Kronecker-Factored Approximations (ICLR 2017 · similarity 0.89)
- On the Parameterization of Second-Order Optimization Effective towards the Infinite Width (ICLR 2024 · similarity 0.89)
- Exact, Tractable Gauss-Newton Optimization in Deep Reversible Architectures Reveal Poor Generalization (NeurIPS 2024 · similarity 0.88)
- Preserving Deep Representations in One-Shot Pruning: A Hessian-Free Second-Order Optimization Framework (ICLR 2025 · similarity 0.86)
- Sketched Adaptive Distributed Deep Learning: A Sharp Convergence Analysis (NeurIPS 2025 · similarity 0.86)
- Convolutions and More as Einsum: A Tensor Network Perspective with Advances for Second-Order Methods (NeurIPS 2024 · 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.