Research map / Architectures and efficiency / RNNs and normalization
Continual Learning With Extended Kronecker-Factored Approximate Curvature
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 XK-FAC instead of K-FAC | ResNet-50 continual learning · ImageNet | Top-1 validation accuracy: 50.37 -> 61.15 | Helped |
| Use XK-FAC instead of K-FAC | ResNet-50 continual learning · Birds | Top-1 validation accuracy: 70.49 -> 78.97 | Helped |
| Use XK-FAC instead of K-FAC | ResNet-50 continual learning · Cars | Top-1 validation accuracy: 85.98 -> 85.91 | Hurt |
| Use XK-FAC instead of K-FAC | ResNet-50 continual learning · Flowers | Top-1 validation accuracy: 88.45 -> 91.84 | Helped |
| Add adaptive αs and αt (ablation) | XK-FAC · ImageNet | Top-1 validation accuracy: 61.15 -> 66.57 | Helped |
| Add adaptive αs and αt (ablation) | XK-FAC · Birds | Top-1 validation accuracy: 78.97 -> 80.30 | Helped |
| Add adaptive αs and αt (ablation) | XK-FAC · Cars | Top-1 validation accuracy: 85.91 -> 88.07 | Helped |
| Add adaptive αs and αt (ablation) | XK-FAC · Flowers | Top-1 validation accuracy: 91.84 -> 92.56 | Helped |
| Train with adaptive αs and αt (ablation) | Fixed BN baseline · ImageNet | Top-1 validation accuracy: 29.21 -> 49.57 | Helped |
| Train with adaptive αs and αt (ablation) | K-FAC baseline · ImageNet | Top-1 validation accuracy: 50.37 -> 65.31 | Helped |
| Interpret BN weights as BRN weights when constructing merged weights (ablation) | BN network with XK-FAC · Permuted MNIST | average validation accuracy | Helped |
| Use BRN rather than BN layers (ablation) | XK-FAC continual learning · Long Permuted MNIST task sequences | average validation accuracy | Helped |
Closest papers by meaning
- Kronecker-Factored Approximate Curvature for Modern Neural Network Architectures (NeurIPS 2023 · similarity 0.90)
- Kronecker-factored Curvature Approximations for Recurrent Neural Networks (ICLR 2018 · similarity 0.90)
- SKFAC: Training Neural Networks With Faster Kronecker-Factored Approximate Curvature (CVPR 2021 · similarity 0.88)
- Kronecker-Factored Approximate Curvature for Physics-Informed Neural Networks (NeurIPS 2024 · similarity 0.88)
- KAC: Kolmogorov-Arnold Classifier for Continual Learning (CVPR 2025 · similarity 0.86)
- A Statistical Theory of Regularization-Based Continual Learning (ICML 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.