Parallel Distributed Logistic Regression for Vertical Federated Learning without Third-Party Coordinator

TL;DR

This paper introduces a scalable parallel distributed logistic regression method for vertical federated learning that eliminates the need for a third-party coordinator, enhancing privacy and efficiency.

Contribution

It proposes a novel coordinator-free approach for vertical federated learning using parameter server architecture, improving scalability and training speed.

Findings

System demonstrates high scalability with large datasets

Training speed significantly improved over existing methods

Effective privacy preservation without third-party involvement

Abstract

Federated Learning is a new distributed learning mechanism which allows model training on a large corpus of decentralized data owned by different data providers, without sharing or leakage of raw data. According to the characteristics of data dis-tribution, it could be usually classified into three categories: horizontal federated learning, vertical federated learning, and federated transfer learning. In this paper we present a solution for parallel dis-tributed logistic regression for vertical federated learning. As compared with existing works, the role of third-party coordinator is removed in our proposed solution. The system is built on the pa-rameter server architecture and aims to speed up the model training via utilizing a cluster of servers in case of large volume of training data. We also evaluate the performance of the parallel distributed model training and the experimental results show the great scalability of the system.