English

DARTS: Differentiable Architecture Search

Machine Learning 2019-04-24 v2 Computation and Language Computer Vision and Pattern Recognition Machine Learning

Abstract

This paper addresses the scalability challenge of architecture search by formulating the task in a differentiable manner. Unlike conventional approaches of applying evolution or reinforcement learning over a discrete and non-differentiable search space, our method is based on the continuous relaxation of the architecture representation, allowing efficient search of the architecture using gradient descent. Extensive experiments on CIFAR-10, ImageNet, Penn Treebank and WikiText-2 show that our algorithm excels in discovering high-performance convolutional architectures for image classification and recurrent architectures for language modeling, while being orders of magnitude faster than state-of-the-art non-differentiable techniques. Our implementation has been made publicly available to facilitate further research on efficient architecture search algorithms.

Keywords

Cite

@article{arxiv.1806.09055,
  title  = {DARTS: Differentiable Architecture Search},
  author = {Hanxiao Liu and Karen Simonyan and Yiming Yang},
  journal= {arXiv preprint arXiv:1806.09055},
  year   = {2019}
}

Comments

Published at ICLR 2019; Code and pretrained models available at https://github.com/quark0/darts

R2 v1 2026-06-23T02:39:34.310Z