Saved in:
Bibliographic Details
Main Authors: Hu, Gang, Teng, Yinglei, Wu, Pengfei, Ma, Shijun
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.12591
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914390465314816
author Hu, Gang
Teng, Yinglei
Wu, Pengfei
Ma, Shijun
author_facet Hu, Gang
Teng, Yinglei
Wu, Pengfei
Ma, Shijun
contents Federated learning on heterogeneous edge devices requires personalized compression while preserving aggregation compatibility and stable convergence. We present Curvature-Aware Heterogeneous Federated Pruning (CA-HFP), a practical framework that enables each client perform structured, device-specific pruning guided by a curvature-informed significance score, and subsequently maps its compact submodel back into a common global parameter space via a lightweight reconstruction. We derive a convergence bound for federated optimization with multiple local SGD steps that explicitly accounts for local computation, data heterogeneity, and pruning-induced perturbations; from which a principled loss-based pruning criterion is derived. Extensive experiments on FMNIST, CIFAR-10, and CIFAR-100 using VGG and ResNet architectures under varying degrees of data heterogeneity demonstrate that CA-HFP preserves model accuracy while significantly reducing per-client computation and communication costs, outperforming standard federated training and existing pruning-based baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2603_12591
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle CA-HFP: Curvature-Aware Heterogeneous Federated Pruning with Model Reconstruction
Hu, Gang
Teng, Yinglei
Wu, Pengfei
Ma, Shijun
Machine Learning
Artificial Intelligence
Federated learning on heterogeneous edge devices requires personalized compression while preserving aggregation compatibility and stable convergence. We present Curvature-Aware Heterogeneous Federated Pruning (CA-HFP), a practical framework that enables each client perform structured, device-specific pruning guided by a curvature-informed significance score, and subsequently maps its compact submodel back into a common global parameter space via a lightweight reconstruction. We derive a convergence bound for federated optimization with multiple local SGD steps that explicitly accounts for local computation, data heterogeneity, and pruning-induced perturbations; from which a principled loss-based pruning criterion is derived. Extensive experiments on FMNIST, CIFAR-10, and CIFAR-100 using VGG and ResNet architectures under varying degrees of data heterogeneity demonstrate that CA-HFP preserves model accuracy while significantly reducing per-client computation and communication costs, outperforming standard federated training and existing pruning-based baselines.
title CA-HFP: Curvature-Aware Heterogeneous Federated Pruning with Model Reconstruction
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2603.12591