CA-HFP: Curvature-Aware Heterogeneous Federated Pruning with Model Reconstruction
This addresses the challenge of efficient and stable federated learning for edge devices with heterogeneous resources, representing an incremental improvement over existing pruning-based methods.
The paper tackled the problem of federated learning on heterogeneous edge devices by proposing CA-HFP, a framework for personalized pruning with model reconstruction, which preserved model accuracy while significantly reducing per-client computation and communication costs, outperforming standard federated training and existing baselines in experiments on datasets like CIFAR-10 and CIFAR-100.
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.