AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Data-Driven Sparse Structure Selection for Deep Neural Networks

Zehao Huang, Naiyan Wang

ECCV 2018 · Architectures and efficiency › Pruning and efficient tuning · approach: Sparse training from scratch · 598 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
optimize scaling factors with APG instead of SGD with hard threshold (ablation)SSS ResNet-32 · ImageNet valtop-1 error (%): 26.46 -> 25.82Improved
use APG instead of ISTA (ablation)SSS block pruning · ImageNet valconvergenceImproved
prune ResNet-50 blocks to ResNet-41ResNet-50 · ImageNet valtop-1 error (%): 23.88 -> 24.56Hurt
prune ResNet-50 blocks to ResNet-32ResNet-50 · ImageNet valtop-1 error (%): 23.88 -> 25.82Hurt
prune ResNeXt-50 blocks and groups to ResNeXt-41ResNeXt-50 · ImageNet valtop-1 error (%): 22.43 -> 24.07Hurt
prune PeleeNet neurons to PeleeNet-APeleeNet · ImageNet valtop-1 error (%): 27.47 -> 27.85Hurt
prune conv1 to conv4 of VGG16 with conv5 penalty set to 0VGG16 · ImageNet valtop-1 error (%): 27.54 -> 31.47Hurt
use SSS ResNeXt-38 instead of DenseNet-121ResNeXt-38 · ImageNet valtop-5 error (%): 7.71 -> 7.50Improved
use SSS ResNet-32 instead of pruned ResNet-50 from channel pruningResNet-32 · ImageNet valtop-1 error (%): 27.70 -> 25.82Improved
use SSS ResNet-41 instead of ResNet-101-prunedResNet-41 · ImageNet valtop-1 error (%): 25.44 -> 24.56Improved
apply SSS block and group pruning to ResNeXt-164ResNeXt-164 · CIFAR-10FLOPs and accuracyImproved
apply SSS neuron pruning to VGGVGG on CIFAR · CIFAR-10 and CIFAR-100parameters and FLOPsMixed

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