4 accepted papers
In this paper, we consider the problem of online learning with convex objectives. Most of the existing work shows that static and dynamic regrets scale logarithmically or sub-linearly with the time horizon T. On the contrary, for the strongly convex setting, it is shown that the static and dynamic r…
In this paper, we consider the challenging problem of online learning with non-convex time varying objectives/loss functions. We show that a simple Online gradient descent (OGD) algorithm achieves a dynamic regret comparable to the strongly convex setting provided a new condition that we propose is…
In this paper, we investigate the problem of caching in a single server setting from the stochastic optimization viewpoint. The goal here is to optimize the time average cache hit subject to a time average constraint on the fetching cost and the cache constraint when the demands are non-stationary a…
In this paper, we consider the problem of Federated Learning (FL) under non-i.i.d data setting. We provide an improved estimate of the empirical loss at each node by using a weighted average of losses across nodes with a penalty term. These uneven weights to different nodes are assigned by taking a…