Q
Quoc V. Le
Researcher at Google
Publications - 229
Citations - 127721
Quoc V. Le is an academic researcher from Google. The author has contributed to research in topics: Artificial neural network & Language model. The author has an hindex of 103, co-authored 217 publications receiving 101217 citations. Previous affiliations of Quoc V. Le include Northwestern University & Tel Aviv University.
Papers
More filters
Proceedings Article
Don't decay the learning rate, increase the batch size
TL;DR: In this article, the authors show that by increasing the batch size during training, one can obtain the same learning curve on both training and test sets, but with fewer parameter updates, leading to greater parallelism and shorter training times.
Posted Content
Meta Pseudo Labels
TL;DR: This work presents Meta Pseudo Labels, a semi-supervised learning method that achieves a new state-of-the-art top-1 accuracy of 90.2% on ImageNet, which is 1.6% better than the existing state of the art [16].
Proceedings ArticleDOI
Unsupervised Pretraining for Sequence to Sequence Learning
TL;DR: This article proposed a general unsupervised learning method to improve the accuracy of sequence-to-sequence (seq2seq) models and achieved state-of-the-art results on the WMT English→German task.
Proceedings Article
DropBlock: A regularization method for convolutional networks
TL;DR: DropBlock is introduced, a form of structured dropout, where units in a contiguous region of a feature map are dropped together, and it is found that applying DropbBlock in skip connections in addition to the convolution layers increases the accuracy.
Posted Content
Large-Scale Evolution of Image Classifiers
Esteban Real,Sherry Moore,Andrew Selle,Saurabh Saxena,Yutaka Leon Suematsu,Jie Tan,Quoc V. Le,Alexey Kurakin +7 more
TL;DR: It is shown that it is now possible to evolve models with accuracies within the range of those published in the last year, starting from trivial initial conditions and reaching accuracies of 94.6% and 77.0%, respectively.