Stochastic Constrained Decentralized Optimization for Machine Learning with Fewer Data Oracles: a Gradient Sliding Approach

arXiv:2404.02511v1 Announce Type: cross
Abstract: In modern decentralized applications, ensuring communication efficiency and privacy for the users are the key challenges. In order to train machine-learning models, the algorithm has to communicate to the data center and sample data for its gradient computation, thus exposing the data and increasing the communication cost. This gives rise to the need for a decentralized optimization algorithm that is communication-efficient and minimizes the number of gradient computations. To this end, we propose the primal-dual …

abstract algorithm applications arxiv center challenges communication computation cs.lg data data center decentralized efficiency gradient key machine machine learning math.oc modern optimization privacy sample stochastic the algorithm the key train type