aggregation_strategies.md

Aggregation Strategies

The aggregation strategy defines how the local updates, computed by the devices, are aggregated to form the global model.

FELES provides different aggregation strategies in order to realize the federated algorithms:

name	reference	details
fedavg	McMahan, Brendan, et al. "Communication-efficient learning of deep networks from decentralized data." Artificial intelligence and statistics. PMLR, 2017.	FedAvg
fednova	Wang, Jianyu, et al. "Tackling the objective inconsistency problem in heterogeneous federated optimization." arXiv preprint arXiv:2007.07481 (2020).	FedNova
feddyn	Acar, Durmus Alp Emre, et al. "Federated learning based on dynamic regularization." arXiv preprint arXiv:2111.04263 (2021).	FedDyn

---

FedAvg Aggregation Strategy

The FedAvg Aggregation Strategy computes a weighted average on the number of examples of the client models.

• Reference: Wang, Jianyu, et al. "Tackling the objective inconsistency problem in heterogeneous federated optimization." arXiv preprint arXiv:2007.07481 (2020).
• Algorithm:

FedNova Aggregation Strategy

The FedNova Aggregation Strategy normalizes and scales the local updates of each party according to their number of local steps before updating the global model.

• Reference: Wang, Jianyu, et al. "Tackling the objective inconsistency problem in heterogeneous federated optimization." arXiv preprint arXiv:2007.07481 (2020).
• Algorithm:

FedDyn Aggregation Strategy

The FedDyn Aggregation Strategy computes the average of the client models and subtracts state h.

• Reference: Acar, Durmus Alp Emre, et al. "Federated learning based on dynamic regularization." arXiv preprint arXiv:2111.04263 (2021).
• Algorithm: