Distributed Learning with a Local Touch: Improving Efficiency in Multiparty Learning

Multiparty learning as an emerging topic, many of the related frameworks and ap-plications are proposed. In this section, we explore the extent of these frameworks and technologies. Yang et al.72 provide a comprehensive survey of existing works on a secure fed-erated learning framework. Bonawitz et al.8 build a scalable production system for Federated Learning in the domain of mobile devices. Konečn`yetal.30 propose ways to reduce communication costs in federated learning. Nishio and Yonetani44 propose a new Federated Learning protocol, FedCS, which can actively manage computing workers based on their resource conditions. Zhao et al.75 notice that conventional federated learning fails on learning non-IID data and propose a strategy to improve training on non-IID data by creating a small subset of data which is globally shared between all the edge devices. Smith et al.63 propose fed-erated multi-task learning, which is a novel systems-aware optimization method, MOCHA.