AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Towards Data-Agnostic Pruning At Initialization: What Makes a Good Sparse Mask?

Hoang Pham, The-Anh Ta, Shiwei Liu, Lichuan Xiang, Dung D. Le, Hongkai Wen, Long Tran-Thanh

NeurIPS 2023 · Architectures and efficiency › Pruning and efficient tuning · approach: Structured network pruning

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 NPB node-path balancing instead of SNIPSNIP · VGG-19 CIFAR-100, 99% sparsityaccuracy (%): 61.77 -> 64.82Improved
Use NPB instead of SynFlowSynFlow · ResNet-20 CIFAR-10, 99% sparsityaccuracy (%): 66.05 -> 70.70Improved
Use NPB instead of Iter-SNIPIter-SNIP · ResNet-18 Tiny-ImageNet, 99% sparsityaccuracy (%): 36.42 -> 41.05Improved
Use NPB instead of PHEWPHEW · ResNet-18 Tiny-ImageNet, 90% sparsityaccuracy (%): 55.93 -> 56.82Improved
Use NPB instead of PHEWPHEW · ResNet-20 CIFAR-10, 99% sparsityaccuracy (%): 70.44 -> 70.70No clear effect
Use NPB instead of PHEWPHEW · ResNet-20 CIFAR-10, 96.84% sparsityaccuracy (%): 81.05 -> 80.55No clear effect
Use NPB instead of PHEW for pruning timePHEW · ResNet-18 Tiny-ImageNet, 90% sparsitypruning time (s): 1342.03 -> 375.65Improved
Use NPB instead of SynFlow for pruning timeSynFlow · ResNet-18 Tiny-ImageNet, 90% sparsitypruning time (s): 96.18 -> 375.65Hurt
Use NPB instead of PHEW for inference costPHEW · ResNet-18 Tiny-ImageNet, 99% sparsityFLOPs (1e8): 1.50 -> 0.59Improved
Use NPB instead of SNIPSNIP · Three settings, various sparsitiesaccuracy: up to 14%Improved
Use NPB instead of SynFlowSynFlow · Three settings, various sparsitiesaccuracy: up to 7%Improved
Random pruning at sparsity above 90%Random pruning · Three settingsaccuracyHurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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