AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

DADA: Differentiable Automatic Data Augmentation

Yonggang Li, Guosheng Hu, Yongtao Wang, Timothy Hospedales, Neil M. Robertson, Yongxin Yang

ECCV 2020 · Architectures and efficiency › Neural architecture search · approach: AutoML pipeline search

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
search policy on full CIFAR-10 instead of reduced set (ablation)DADA · CIFAR-10 WRN-40-2test error: 3.61 -> 3.53Improved
search policy on full CIFAR-10 instead of reduced set (ablation)DADA · CIFAR-10 WRN-28-10test error: 2.73 -> 2.64Improved
use unbiased RELAX estimator instead of Gumbel-Softmax estimator (ablation)DADA · CIFAR-10test errorImproved
apply DADA policy versus no augmentation policyWRN-40-2 · CIFAR-10test error: 5.3 -> 3.6Improved
apply DADA policy versus no augmentation policyResNet-50 · ImageNettop-1 error: 23.7 -> 22.5Improved
pretrain backbone with DADA policyRetinaNet ResNet-50 · COCObox AP: 35.9 -> 36.6Improved
pretrain backbone with DADA policyFaster R-CNN ResNet-50 · COCObox AP: 36.6 -> 37.2Improved
pretrain backbone with DADA policyMask R-CNN ResNet-50 · COCOmask AP: 34.2 -> 34.5Improved
DADA instead of Fast AA for searchpolicy search · CIFAR-10GPU hours: 3.5 -> 0.1Improved
DADA versus AA on CIFAR-100 WRN-28-10policy search · CIFAR-100test error: 17.1 -> 17.5Hurt
DADA versus Fast AA on full-dataset searchpolicy search · CIFAR-10test errorMixed

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