Unsupervised Data Augmentation for Consistency Training

arXiv 2019 Qizhe XieZihang DaiEduard HovyMinh-Thang LuongQuoc V. Le

Despite much success, deep learning generally does not perform well with small labeled training sets. In these scenarios, data augmentation has shown much promise in alleviating the need for more labeled data, but it so far has mostly been applied in supervised settings and achieved limited gains... (read more)

PDF Abstract
Task Dataset Model Metric name Metric value Global rank Compare
Text Classification Amazon-2 BERT Finetune + UDA Error 3.50 # 2
Text Classification Amazon-5 BERT Finetune + UDA Error 37.12 # 2
Semi-Supervised Image Classification ImageNet - 10% labeled data Unsup. Data Augmentation Top 5 Accuracy 88.52% # 2
Text Classification IMDb BERT Finetune + UDA Accuracy 95.80 # 2
Text Classification Yelp-2 BERT Finetune + UDA Accuracy 97.95% # 2
Text Classification Yelp-5 BERT Finetune + UDA Accuracy 67.92% # 2