AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Fast Differentiable Matrix Square Root

Yue Song, Nicu Sebe, Wei Wang

ICLR 2022 · Architectures and efficiency › RNNs and normalization · approach: Weight normalization and decay interplay · 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 MPA-Lya instead of NS iterationmatrix square root computation · single matrices from 256×256 to 1024×1024time (ms): 6.28 -> 5.56; 9.86 -> 7.47; 12.69 -> 11.28Improved
Use MTP-Lya instead of NS iterationmatrix square root computation · single matrices from 256×256 to 1024×1024time (ms): 6.28 -> 4.88; 9.86 -> 5.43; 12.69 -> 7.32Improved
Use MPA-Lya instead of NS iterationResNet-50 global covariance pooling · ImageNet valtime (ms), single FP+BP step: 164.43 -> 110.61Improved
Use MPA-Lya instead of NS iterationResNet-50 global covariance pooling · ImageNet valtop-1 accuracy (%): 77.19 -> 77.13No clear effect
Use MPA-Lya instead of NS iterationResNet-50 global covariance pooling · ImageNet valtop-5 accuracy (%): 93.40 -> 93.45No clear effect
Use MPA-Lya instead of NS iterationZCA whitening · 1×64×64 matrixFP+BP time (ms): 2.96 -> 2.52Improved
Use MTP-Lya instead of NS iterationZCA whitening · 1×64×64 matrixFP+BP time (ms): 2.96 -> 2.36Improved
Use MPA-Lya instead of NS iterationSo-ViT · batch 64 of 48×48 matricesFP+BP time (ms): 10.38 -> 3.25Improved
Use MTP-Lya instead of NS iterationSo-ViT · batch 64 of 48×48 matricesFP+BP time (ms): 10.38 -> 2.39Improved
Use the Huang et al. NS iteration instead of the Higham NS iterationNS iteration · ResNet-18 on CIFAR10 and CIFAR100mean accuracy: 4.57±0.15 vs 4.53±0.14; 21.24±0.20 vs 21.28±0.17No clear effect
Use the Huang et al. NS iteration instead of the Higham NS iterationNS iteration · ResNet-50 on CIFAR100mean accuracy: 19.39±0.30 -> 19.53±0.35Hurt
Increase Lyapunov solver iterations from 5 to 9 (ablation)backward Lyapunov solver · 10,000 randomly sampled covariance matrices||B_k−I||_F estimation error: 0.3541 -> 2e-7Improved

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