Coded Computing

Coded computing enhances distributed computing by encoding data before distributing it to worker nodes, enabling fault tolerance and improved efficiency in the face of slow or faulty servers (stragglers). Current research focuses on adapting this framework to machine learning tasks, particularly federated learning, using techniques like polynomial interpolation and regression models to efficiently decode results and maintain data privacy. This approach offers significant potential for accelerating large-scale computations and improving the resilience and security of distributed systems, impacting areas such as AI model training and prediction serving.

Papers