Efficient Architecture Search by Network Transformation

TL;DR

This paper introduces an efficient neural architecture search method that reuses network weights during exploration, significantly reducing computational costs while designing competitive CNNs on image benchmarks.

Contribution

It proposes a reinforcement learning framework that employs function-preserving transformations to grow networks, enabling weight reuse and efficient architecture search.

Findings

Achieved 4.23% test error on CIFAR-10 without skip-connections.

Designed networks outperform existing models with similar schemes.

Explored DenseNet space to find more accurate, parameter-efficient networks.

Abstract

Techniques for automatically designing deep neural network architectures such as reinforcement learning based approaches have recently shown promising results. However, their success is based on vast computational resources (e.g. hundreds of GPUs), making them difficult to be widely used. A noticeable limitation is that they still design and train each network from scratch during the exploration of the architecture space, which is highly inefficient. In this paper, we propose a new framework toward efficient architecture search by exploring the architecture space based on the current network and reusing its weights. We employ a reinforcement learning agent as the meta-controller, whose action is to grow the network depth or layer width with function-preserving transformations. As such, the previously validated networks can be reused for further exploration, thus saves a large amount of computational cost. We apply our method to explore the architecture space of the plain convolutional neural networks (no skip-connections, branching etc.) on image benchmark datasets (CIFAR-10, SVHN) with restricted computational resources (5 GPUs). Our method can design highly competitive networks that outperform existing networks using the same design scheme. On CIFAR-10, our model without skip-connections achieves 4.23\% test error rate, exceeding a vast majority of modern architectures and approaching DenseNet. Furthermore, by applying our method to explore the DenseNet architecture space, we are able to achieve more accurate networks with fewer parameters.