Federated Neural Architecture Search

TL;DR

This paper introduces FedNAS, an efficient federated neural architecture search framework that enables privacy-preserving model design for decentralized data, achieving high accuracy with reduced client costs.

Contribution

It proposes FedNAS, a novel federated NAS method optimized for limited client resources, with key innovations like parallel training, early dropping, and dynamic rounds.

Findings

Achieves comparable accuracy to centralized NAS methods.

Reduces client computational and communication costs significantly.

Demonstrates effectiveness on large-scale datasets and CNN architectures.

Abstract

To preserve user privacy while enabling mobile intelligence, techniques have been proposed to train deep neural networks on decentralized data. However, training over decentralized data makes the design of neural architecture quite difficult as it already was. Such difficulty is further amplified when designing and deploying different neural architectures for heterogeneous mobile platforms. In this work, we propose an automatic neural architecture search into the decentralized training, as a new DNN training paradigm called Federated Neural Architecture Search, namely federated NAS. To deal with the primary challenge of limited on-client computational and communication resources, we present FedNAS, a highly optimized framework for efficient federated NAS. FedNAS fully exploits the key opportunity of insufficient model candidate re-training during the architecture search process, and incorporates three key optimizations: parallel candidates training on partial clients, early dropping candidates with inferior performance, and dynamic round numbers. Tested on large-scale datasets and typical CNN architectures, FedNAS achieves comparable model accuracy as state-of-the-art NAS algorithm that trains models with centralized data, and also reduces the client cost by up to two orders of magnitude compared to a straightforward design of federated NAS.