FedEAS policy cuts synthetic data generation costs in federated learning

Researchers have developed FedEAS, a novel policy designed to mitigate label skew in federated learning (FL). This approach optimizes the synthetic data generation process by assigning each client a budget based on its local label distribution, determining both the quantity and destination of generated samples. FedEAS significantly reduces the overall computation cost associated with full class balancing, achieving substantial accuracy gains while using up to 94.1% less generation budget compared to traditional methods. When compared to uniform allocation strategies at the same generation budget, FedEAS demonstrates superior performance, improving accuracy by up to 18.82% on datasets like CIFAR-10 and CIFAR-100. AI

IMPACT Optimizes synthetic data generation in federated learning, potentially reducing computational costs and improving model accuracy.

RANK_REASON The cluster contains a research paper detailing a new method for federated learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• CIFAR-10
• CIFAR-100
• FedEAS
• federated learning

• paper
• infra

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