Optimal MIMO Combining for Blind Federated Edge Learning with Gradient Sparsification

Autor:	Becirovic, Ema, Chen, Zheng, Larsson, Erik G.
Rok vydání:	2022
Předmět:	Computer Science - Information Theory Electrical Engineering and Systems Science - Signal Processing
Druh dokumentu:	Working Paper
Popis:	We provide the optimal receive combining strategy for federated learning in multiple-input multiple-output (MIMO) systems. Our proposed algorithm allows the clients to perform individual gradient sparsification which greatly improves performance in scenarios with heterogeneous (non i.i.d.) training data. The proposed method beats the benchmark by a wide margin. Comment: 5 pages, 1 figure, accepted to IEEE SPAWC 2022
Databáze:	arXiv
Externí odkaz:	http://arxiv.org/abs/2203.12957 Zobrazit plný text záznamu View this record from Arxiv