What type of study is this?

This is a Experimental Study study.

Adaptive Batch Size Time Evolving Stochastic Gradient Descent for Federated Learning

View Full Paper

Ask AI

Bookmark

The proposed FedATEVR algorithm enhances the speed of federated learning with a linear speedup of O(1/√SKT).
Using historical gradient information, FedATEVR optimizes mega-batch sizes for clients, improving local SGD efficiency.
The method effectively tackles gradient heterogeneity, stabilizing batch size variations during training.
Extensive experiments confirm that FedATEVR outperforms baseline methods in test accuracy while reducing communication rounds.

No takes yet. Share an insight, caveat, or question.

Proposed algorithm improves accuracy and communication efficiency in federated learning, addressing gradient heterogeneity.

An et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark