A Study of Split Learning Model

Abstract

Split learning is considered a state-of-the-art solution for machine learning privacy that takes place between clients and servers. In this way, the model is split and trained, so that the original data does not move to the client from the server, and the model is properly split between the client and the server, reducing the burden of training. This paper introduces the concept of split learning, reviews traditional, novel, and state-of-the-art split learning methods, and discusses current challenges and trends.