AI Paper Atlas

Research map / Learning theory and algorithms / Matrix and tensor methods

M-FAC: Efficient Matrix-Free Approximations of Second-Order Information

Elias Frantar, Eldar Kurtic, Dan Alistarh

NeurIPS 2021 · Learning theory and algorithms › Matrix and tensor methods · approach: Geometry of low-rank and SPD structure · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (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 changedWhereMeasuredEffect
Use M-FAC instead of WoodFisher for gradual pruningMobileNetV1-STR · ImageNet, 89% sparsityTop-1 accuracy: 63.87 -> 65.01Improved
Use M-FAC instead of WoodFisher for gradual pruningResNet50 · ImageNet, 95% sparsityTop-1 accuracy: 72.12 -> 72.32Improved
Use practical M-FAC settings rather than WoodFisherMobileNetV1 · ImageNet, 89% sparsityaccuracy: 63.9 -> 67.3Improved
Use practical M-FAC settings rather than WoodFisherResNet50 · ImageNet, 98% sparsityaccuracy: 65.6 -> 67.5Improved
Update remaining weights after pruning instead of leaving them unchanged (ablation)M-FAC pruning · ResNet20 / CIFAR-10, 90% sparsityaccuracy: 11.5 -> 28.2Improved
Use approximate simultaneous OBS updates rather than solving the full linear system (ablation)M-FAC pruning · ResNet20 / CIFAR-10, 90% sparsityaccuracy: 28.1 vs 28.2No clear effect
Increase the number of gradients used for Fisher estimation (ablation)M-FAC one-shot pruning · ResNet50 / ImageNetpruned accuracyImproved
Increase the number of gradients used for Fisher estimation (ablation)M-FAC one-shot pruning · MobileNetV1-STR / ImageNetpruned accuracyImproved
Increase block size while holding the gradient count fixed (ablation)M-FAC one-shot pruning · ResNet50 / ImageNetpruned accuracyMixed
Increase block size while holding the gradient count fixed (ablation)M-FAC one-shot pruning · MobileNetV1-STR / ImageNetpruned accuracyHurt
Use M-FAC instead of SGDResNet20 · CIFAR-10test accuracy: 91.78 -> 92.34Improved
Use M-FAC instead of SGDResNet32 · CIFAR-10test accuracy: 92.80 -> 92.65No clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy