Adaptive client selection in resource constrained federated learning systems

The Adaptive client selection using federated learning is a privacy-preserving Federated Learning framework designed to decentralize machine learning by training models directly on mobile devices. By keeping raw data on-device and only transmitting model updates to a central server, the project ensures user anonymity and data security. A core innovation of this system is its Hardware-Aware Client Selection mechanism. It allows users to input specific hardware constraints-including minimum RAM, storage (GB), and processor type and processes these parameters to identify and output the optimal client device for the training task. This ensures high model accuracy while maintaining system efficiency across diverse mobile hardware.