学习可扩展图像识别的可转换架构.pdf

Learning Transferable Architectures for Scalable Image Recognition 8 Abstract 1. Introduction 1 0 2 Developing neural network image classification models often requires significant architecture engineering. Starting r p Developing neural network image classification models from the seminal work of [32] on using convolutional archi- A often requires significant architecture engineering. In this tectures [17, 34] for ImageNet [11] classification, succes- paper, we study a method to learn the model architectures sive advancements through architecture engineering have 1 directly on the dataset of interest. As this approach is ex- achieved impressive results [53, 59, 20, 60, 58, 68]. 1 pensive when the dataset is large, we propose to search for In this paper, we study a new paradigm of designing con- ] an architectural building block on a small dataset and then volutional architectures and describe a scalable method to V transfer the block to a larger dataset. The key contribu- optimize convolutional architectures on a dataset of inter- C tion of this work is the design of a new search space (which est, for instance the ImageNet classification dataset. Our s. we call the “NASNet search space”) which enables trans- approach is inspired by the recently proposed Neural Ar- c ferability. In our experiments, we search for the best con- chitecture Search (NAS) framework [71], which uses a re- [ volutional layer (or “cell”) on the CIFAR-10 dataset and inforcement learning search method to optimize architec- 4 then apply this cell to the