Neural Architecture Search in Embedding Space

TL;DR

This paper introduces NASES, a novel neural architecture search framework that searches in an embedding space using reinforcement learning, significantly improving efficiency and performance in image classification tasks.

Contribution

NASES enables reinforcement learning to operate in an embedding space for NAS, overcoming high-dimensional search space challenges and reducing search time.

Findings

NASES achieves comparable performance to other NAS methods on CIFAR-10.

NASES discovers architectures in less than 3.5 GPU hours.

Embedding-based search improves efficiency and effectiveness.

Abstract

The neural architecture search (NAS) algorithm with reinforcement learning can be a powerful and novel framework for the automatic discovering process of neural architectures. However, its application is restricted by noncontinuous and high-dimensional search spaces, which result in difficulty in optimization. To resolve these problems, we proposed NAS in embedding space (NASES), which is a novel framework. Unlike other NAS with reinforcement learning approaches that search over a discrete and high-dimensional architecture space, this approach enables reinforcement learning to search in an embedding space by using architecture encoders and decoders. The current experiment demonstrated that the performance of the final architecture network using the NASES procedure is comparable with that of other popular NAS approaches for the image classification task on CIFAR-10. The results of the experiment were efficient and indicated that NASES was highly efficient to discover final architecture only in $<$3.5 GPU hours. The beneficial-performance and effectiveness of NASES was impressive when the architecture-embedding searching and weight initialization were applied.