Asymptotic Behaviors of Projected Stochastic Approximation: A Jump Diffusion Perspective

Neural Information Processing Systems 

In this paper we consider linearly constrained stochastic approximation problems with federated learning as a special case.