AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

You Only Search Once: Single Shot Neural Architecture Search via Direct Sparse Optimization

Xinbang Zhang, Zehao Huang, Naiyan Wang, Shiming Xiang, Chunhong Pan

TPAMI 2020 · Architectures and efficiency › Neural architecture search · approach: Efficient NAS search strategies · 95 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
add split training sets for weights and structure (ablation)DSO-NAS-full · CIFAR-10test error (%): 3.20 -> 3.02Improved
add split training sets for weights and structure (ablation)DSO-NAS-share · CIFAR-10test error (%): 3.14 -> 2.86Improved
add pre-trained weight initialization before search (ablation)DSO-NAS-full · CIFAR-10test error (%): 3.26 -> 3.02Improved
add pre-trained weight initialization before search (ablation)DSO-NAS-share · CIFAR-10test error (%): 3.07 -> 2.86Improved
change weight:structure update ratio from 1:1 to 4:1 (ablation)DSO-NAS-full · CIFAR-10test error (%): 3.02 -> 3.05No clear effect
change weight:structure update ratio from 1:1 to 4:1 (ablation)DSO-NAS-share · CIFAR-10test error (%): 2.86 -> 2.89No clear effect
use block sharing (share) instead of full search spaceDSO-NAS · CIFAR-10 (with cutout)test error (%): 2.95 -> 2.84No clear effect
use DSO-NAS searched structure instead of random structureDSO-NAS-share · CIFAR-10 (with cutout)test error (%): 3.58 -> 2.84Improved
search directly on ImageNet instead of transferring CIFAR-10 blockDSO-NAS · ImageNettop-1 error (%): 26.2 -> 25.4Improved
add adaptive FLOPs regularization (ablation)DSO-NAS · CIFAR-10test errorImproved
add adaptive MAC regularization (ablation)DSO-NAS · CIFAR-10accuracy under MAC budgetImproved

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