Comparison of Privacy-Preserving Distributed Deep Learning Methods in Healthcare

TL;DR

This paper compares three privacy-preserving distributed deep learning methods in healthcare, introduces a new architecture called SplitFedv3, and proposes an improved training technique for better performance and efficiency.

Contribution

It introduces SplitFedv3, a novel distributed learning architecture, and a new training method called alternate mini-batch training, enhancing performance and efficiency.

Findings

SplitFedv3 outperforms split learning and SplitFedv2 in experiments.

Alternate mini-batch training improves over alternate client training.

The methods effectively balance privacy, performance, and computational costs.

Abstract

In this paper, we compare three privacy-preserving distributed learning techniques: federated learning, split learning, and SplitFed. We use these techniques to develop binary classification models for detecting tuberculosis from chest X-rays and compare them in terms of classification performance, communication and computational costs, and training time. We propose a novel distributed learning architecture called SplitFedv3, which performs better than split learning and SplitFedv2 in our experiments. We also propose alternate mini-batch training, a new training technique for split learning, that performs better than alternate client training, where clients take turns to train a model.