VARS-FL builds client reputation from validation loss reduction signals and uses sliding-window averaging plus log-scaled participation to select clients, yielding up to 36% faster convergence to 80% accuracy on non-IID Edge-IIoTset intrusion detection.
Internet of things intrusion detection: Centralized, on-device, or feder- ated learning?
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
VARS-FL: Validation-Aligned Client Selection for Non-IID Federated Learning in IoT Systems
VARS-FL builds client reputation from validation loss reduction signals and uses sliding-window averaging plus log-scaled participation to select clients, yielding up to 36% faster convergence to 80% accuracy on non-IID Edge-IIoTset intrusion detection.