Cross-client Label Propagation for Transductive and Semi-Supervised Federated Learning

TL;DR

This paper introduces Cross-Client Label Propagation (XCLP), a secure federated learning method that propagates labels across data graphs from multiple clients, improving classification accuracy in transductive and semi-supervised settings.

Contribution

XCLP is a novel secure graph-based label propagation method for federated learning, enabling label estimation without data sharing and enhancing accuracy in various applications.

Findings

XCLP outperforms alternative methods in classification accuracy.

XCLP effectively predicts labels for unseen test data.

XCLP improves semi-supervised federated learning performance.

Abstract

We present Cross-Client Label Propagation(XCLP), a new method for transductive federated learning. XCLP estimates a data graph jointly from the data of multiple clients and computes labels for the unlabeled data by propagating label information across the graph. To avoid clients having to share their data with anyone, XCLP employs two cryptographically secure protocols: secure Hamming distance computation and secure summation. We demonstrate two distinct applications of XCLP within federated learning. In the first, we use it in a one-shot way to predict labels for unseen test points. In the second, we use it to repeatedly pseudo-label unlabeled training data in a federated semi-supervised setting. Experiments on both real federated and standard benchmark datasets show that in both applications XCLP achieves higher classification accuracy than alternative approaches.