Heterogeneity-Aware Personalized Federated Neural Architecture Search

Federated learning (FL), which enables collaborative learning across distributed nodes, confronts a significant heterogeneity challenge, primarily including resource heterogeneity induced by different hardware platforms, and statistical heterogeneity originating from non-IID private data distributio...

Full description

Saved in:
Bibliographic Details
Main Authors: An Yang, Ying Liu
Format: Article
Language:English
Published: MDPI AG 2025-07-01
Series:Entropy
Subjects:
Online Access:https://www.mdpi.com/1099-4300/27/7/759
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849733982514577408
author An Yang
Ying Liu
author_facet An Yang
Ying Liu
author_sort An Yang
collection DOAJ
description Federated learning (FL), which enables collaborative learning across distributed nodes, confronts a significant heterogeneity challenge, primarily including resource heterogeneity induced by different hardware platforms, and statistical heterogeneity originating from non-IID private data distributions among clients. Neural architecture search (NAS), particularly one-shot NAS, holds great promise for automatically designing optimal personalized models tailored to such heterogeneous scenarios. However, the coexistence of both resource and statistical heterogeneity destabilizes the training of the one-shot supernet, impairs the evaluation of candidate architectures, and ultimately hinders the discovery of optimal personalized models. To address this problem, we propose a heterogeneity-aware personalized federated NAS (HAPFNAS) method. First, we leverage lightweight knowledge models to distill knowledge from clients to server-side supernet, thereby effectively mitigating the effects of heterogeneity and enhancing the training stability. Then, we build random-forest-based personalized performance predictors to enable the efficient evaluation of candidate architectures across clients. Furthermore, we develop a model-heterogeneous FL algorithm called heteroFedAvg to facilitate collaborative model training for the discovered personalized models. Comprehensive experiments on CIFAR-10/100 and Tiny-ImageNet classification datasets demonstrate the effectiveness of our HAPFNAS, compared to state-of-the-art federated NAS methods.
format Article
id doaj-art-58cfc318f152402baf6dfd215f7013a1
institution DOAJ
issn 1099-4300
language English
publishDate 2025-07-01
publisher MDPI AG
record_format Article
series Entropy
spelling doaj-art-58cfc318f152402baf6dfd215f7013a12025-08-20T03:07:55ZengMDPI AGEntropy1099-43002025-07-0127775910.3390/e27070759Heterogeneity-Aware Personalized Federated Neural Architecture SearchAn Yang0Ying Liu1College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, ChinaCollege of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, ChinaFederated learning (FL), which enables collaborative learning across distributed nodes, confronts a significant heterogeneity challenge, primarily including resource heterogeneity induced by different hardware platforms, and statistical heterogeneity originating from non-IID private data distributions among clients. Neural architecture search (NAS), particularly one-shot NAS, holds great promise for automatically designing optimal personalized models tailored to such heterogeneous scenarios. However, the coexistence of both resource and statistical heterogeneity destabilizes the training of the one-shot supernet, impairs the evaluation of candidate architectures, and ultimately hinders the discovery of optimal personalized models. To address this problem, we propose a heterogeneity-aware personalized federated NAS (HAPFNAS) method. First, we leverage lightweight knowledge models to distill knowledge from clients to server-side supernet, thereby effectively mitigating the effects of heterogeneity and enhancing the training stability. Then, we build random-forest-based personalized performance predictors to enable the efficient evaluation of candidate architectures across clients. Furthermore, we develop a model-heterogeneous FL algorithm called heteroFedAvg to facilitate collaborative model training for the discovered personalized models. Comprehensive experiments on CIFAR-10/100 and Tiny-ImageNet classification datasets demonstrate the effectiveness of our HAPFNAS, compared to state-of-the-art federated NAS methods.https://www.mdpi.com/1099-4300/27/7/759neural architecture searchneural networkfederated learningpersonalization
spellingShingle An Yang
Ying Liu
Heterogeneity-Aware Personalized Federated Neural Architecture Search
Entropy
neural architecture search
neural network
federated learning
personalization
title Heterogeneity-Aware Personalized Federated Neural Architecture Search
title_full Heterogeneity-Aware Personalized Federated Neural Architecture Search
title_fullStr Heterogeneity-Aware Personalized Federated Neural Architecture Search
title_full_unstemmed Heterogeneity-Aware Personalized Federated Neural Architecture Search
title_short Heterogeneity-Aware Personalized Federated Neural Architecture Search
title_sort heterogeneity aware personalized federated neural architecture search
topic neural architecture search
neural network
federated learning
personalization
url https://www.mdpi.com/1099-4300/27/7/759
work_keys_str_mv AT anyang heterogeneityawarepersonalizedfederatedneuralarchitecturesearch
AT yingliu heterogeneityawarepersonalizedfederatedneuralarchitecturesearch