<ul><li>Byzantine Resilient Federated Multi-Task Representation Learning.</li><li>The paper introduces BR-MTRL, a Byzantine-resilient multi-task representation learning framework for handling faulty or malicious agents.</li><li>The approach utilizes a shared neural network model with fixed layers shared among clients, except for a client-specific final layer.</li><li>The method demonstrates personalized learning while maintaining resilience in distributed settings by employing an alternating gradient descent strategy and geometric median aggregation.</li></ul>

Byzantine Resilient Federated Multi-Task Representation Learning

Discover more