FedCSGA: Client Selection Based on Genetic Algorithm for Federated Learning
What it is
FedCSGA is a client selection method for Federated Learning that uses a Genetic Algorithm to optimize the subset of clients participating in each training round. It encodes client subsets as chromosomes and evaluates them using a fitness function that jointly considers resource-aware metrics and historical participation weights to balance statistical utility with system efficiency.
The problem it solves
Existing client selection methods in Federated Learning often treat statistical heterogeneity (data distribution) and system heterogeneity (resource availability) independently, failing to capture their coupled and dynamic impact on training efficiency and convergence.
What you could use it for
- Optimizing communication costs in heterogeneous edge computing environments
- Improving convergence speed in Federated Learning with non-IID data
- Mitigating straggler effects in distributed model training
Why it matters
If effective, it could significantly reduce the communication overhead and training time of Federated Learning systems in real-world, heterogeneous deployments while improving final model accuracy.
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.