AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

Online Hyper-Parameter Learning for Auto-Augmentation Strategy

Chen Lin, Minghao Guo, Chuming Li, Xin Yuan, Wei Wu, Junjie Yan, Dahua Lin, Wanli Ouyang

ICCV 2019 · Architectures and efficiency › Neural architecture search · approach: AutoML pipeline search · 98 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 OHL-Auto-Aug instead of baseline augmentationResNet-18 · CIFAR-10 testerror rate (%): 4.66 -> 3.29Improved
use OHL-Auto-Aug instead of Auto-AugmentWideResNet-28-10 · CIFAR-10 testerror rate (%): 2.68 -> 2.61Improved
use OHL-Auto-Aug instead of Auto-AugmentAmoebaNet-B(6,128) · CIFAR-10 testerror rate (%): 1.75 -> 1.89Hurt
use OHL-Auto-Aug instead of our-implementation baselineAmoebaNet-B(6,128) · CIFAR-10 testerror rate (%): 3.4 -> 1.89Improved
use OHL-Auto-Aug instead of Auto-AugmentResNet-50 · ImageNet valtop-1 error (%): 22.37 -> 21.07Improved
use OHL-Auto-Aug instead of baselineSE-ResNeXt-101 · ImageNet valtop-1 error (%): 20.70 -> 19.30Improved
use OHL-Auto-Aug instead of Auto-AugmentSE-ResNeXt-101 · ImageNet valtop-5 error (%): 5.23 -> 4.57Hurt
replace Auto-Augment search with online learningsearch cost · CIFAR-10 search costsearch iterations: 7.03x10^6 -> 1.17x10^5Improved
replace Auto-Augment search with online learningsearch cost · ImageNet search costsearch iterations: 1.76x10^7 -> 7.5x10^5Improved
learn augmentation distribution (eta>0) instead of uniform sampling (eta=0) (ablation)OHL-Auto-Aug ResNet-18 · CIFAR-10 testaccuracyImproved
increase trajectory samples from 1 to 8 (ablation)OHL-Auto-Aug ResNet-18 · CIFAR-10 testaccuracyImproved
increase trajectory samples beyond 8 (ablation)OHL-Auto-Aug ResNet-18 · CIFAR-10 testaccuracyNo 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