Federated Learning (FL) is transforming machine learning by enabling the training of powerful AI models without centralizing data. This tutorial offers a comprehensive exploration of FL, beginning with its fundamental principles and the foundational FedAvg algorithm, which aggregates locally trained models to build a global one.
We then examine FL from an economic perspective, discussing how major industry players leverage FL for competitive advantage—sometimes even employing adversarial strategies. Next, we address the challenge of Byzantine workers—malicious or faulty nodes that can disrupt training—and explore robust defense mechanisms against such threats.
The tutorial also introduces Personalized Federated Learning, focusing on strategies for identifying and collaborating with agents that share similar data distributions to improve model performance. Finally, we analyze key bottlenecks in FL, including communication overhead and privacy concerns, and present cutting-edge techniques to overcome these challenges.
By the end of this tutorial, participants will have a deep understanding of Federated Learning’s core concepts, economic implications, security challenges, and recent advancements, equipping them to apply these insights to their research and practical applications.