Advancing Federated Learning with Non-IID Data

This research explores federated learning, emphasizing its privacy-preserving operational process and comparing it with centralized learning. Challenges like non-IID data and imbalanced distributions are analyzed. Key algorithms such as FedAvg and FedProx are discussed, alongside testing considerations like model architecture and simulation setups. Results highlight improved accuracy and hybrid methodologies' advantages. Limitations and future advancements focus on refining parameters,...