AutoDropout: Learning Dropout Patterns to Regularize Deep Networks

Open paper
Benchmark Model Rank Results
image-classification-on-cifar-10WRN-28-10+AutoDropout+RandAugment#64Percentage correct: 97.9
image-classification-on-cifar-10AutoDropout#98Percentage correct: 96.8
image-classification-on-imagenetResNet-50+AutoDropout+RandAugment#675Top 1 Accuracy: 80.3%
image-classification-on-imagenetResNet-50#765Top 1 Accuracy: 78.7%
image-classification-on-imagenetEfficientNet-B0#821Top 1 Accuracy: 77.5%
language-modelling-on-penn-treebank-wordTransformer-XL + AutoDropout#22Test perplexity: 54.9Validation perplexity: 58.1
machine-translation-on-iwslt2014-germanTransformerBase + AutoDropout#17BLEU score: 35.8
machine-translation-on-wmt2014-english-frenchTransformerBase + AutoDropout#34BLEU score: 40

Rank counts only results with a code link.