Gradient-Driven Rewards to Guarantee Fairness in Collaborative Machine Learning

An important distinction of our work here from the standard FL literature is that the agents are self-interested and hence not necessarily cooperative like the worker nodes in distributed learning.