Neural Architecture Search via Combinatorial Multi-Armed Bandit

TL;DR

This paper introduces a novel approach to neural architecture search by formulating it as a combinatorial multi-armed bandit problem, enabling more efficient tree-search methods that achieve competitive results with significantly less computational cost.

Contribution

It proposes a new formulation of NAS as a CMAB problem and applies Nested Monte-Carlo Search, improving efficiency and effectiveness of tree-search methods in NAS.

Findings

Achieves state-of-the-art accuracy on CIFAR-10 with only 0.58 GPU days.

Discovered architecture transfers well to ImageNet.

20x faster than existing tree-search NAS methods.

Abstract

Neural Architecture Search (NAS) has gained significant popularity as an effective tool for designing high performance deep neural networks (DNNs). NAS can be performed via policy gradient, evolutionary algorithms, differentiable architecture search or tree-search methods. While significant progress has been made for both policy gradient and differentiable architecture search, tree-search methods have so far failed to achieve comparable accuracy or search efficiency. In this paper, we formulate NAS as a Combinatorial Multi-Armed Bandit (CMAB) problem (CMAB-NAS). This allows the decomposition of a large search space into smaller blocks where tree-search methods can be applied more effectively and efficiently. We further leverage a tree-based method called Nested Monte-Carlo Search to tackle the CMAB-NAS problem. On CIFAR-10, our approach discovers a cell structure that achieves a low error rate that is comparable to the state-of-the-art, using only 0.58 GPU days, which is 20 times faster than current tree-search methods. Moreover, the discovered structure transfers well to large-scale datasets such as ImageNet.