FANS framework optimizes model architectures for heterogeneous federated learning

Researchers have developed FANS (Federated Adaptive Network Search), a new framework designed to optimize model architectures in heterogeneous federated learning environments. This approach utilizes a hypernetwork to learn a shared architecture space, moving beyond predefined model menus. To enhance efficiency, the Federated Parallel Scaling (FPS) algorithm is introduced, enabling parallel training of multiple subnetworks with self-distillation. Evaluations on CIFAR-10, CIFAR-100, and MNLI datasets demonstrated that FANS significantly expands the range of feasible subnetworks and improves the accuracy-efficiency trade-off compared to existing methods. AI

IMPACT Enhances efficiency and architectural flexibility in federated learning, potentially enabling more complex models on diverse devices.

RANK_REASON Academic paper detailing a new framework and algorithm for federated learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Alireza Furutanpey
• Bert
• CIFAR-10
• CIFAR-100
• DenseNet 121
• FANS
• Federated Adaptive Network Search
• Federated Parallel Scaling
• MnlI
• ResNet-18

• paper
• model release
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →