Federated Learning

Chapter 11: Vertical asynchronous federated learning: algorithms and theoretic guarantees

Tianyi Chen^a; Xiao Jin^a; Yuejiao Sun^b; Wotao Yin^c    aRensselaer Polytechnic Institute, Troy, NY, United States^bUniversity of California, Los Angeles, CA, United States^cAlibaba Group, Bellevue, WA, United States

Abstract

Federated learning (FL) enables model training with multi-client data where privacy concerns and coordination challenges prevent us from using conventional machine learning methods. In particular, horizontal FL handles multi-client data that share the same set of features, and vertical FL handles data of the same set of subjects but different clients hold their different features. This chapter formulates vertical FL as an optimization problem ...