FedSEAL: Semi-Supervised Federated Learning with Self-Ensemble Learning and Negative Learning

TL;DR

FedSEAL introduces a semi-supervised federated learning approach that leverages self-ensemble and negative learning techniques to effectively utilize unlabeled client data and small labeled server data, improving accuracy and efficiency.

Contribution

The paper proposes FedSEAL, a novel semi-supervised federated learning algorithm combining self-ensemble and negative learning, addressing practical scenarios with limited labeled data.

Findings

Outperforms state-of-the-art SSFL methods on Fashion-MNIST and CIFAR10.

Effectively utilizes unlabeled client data and small server-labeled data.

Enhances accuracy and training efficiency in semi-supervised federated learning.

Abstract

Federated learning (FL), a popular decentralized and privacy-preserving machine learning (FL) framework, has received extensive research attention in recent years. The majority of existing works focus on supervised learning (SL) problems where it is assumed that clients carry labeled datasets while the server has no data. However, in realistic scenarios, clients are often unable to label their data due to the lack of expertise and motivation while the server may host a small amount of labeled data. How to reasonably utilize the server labeled data and the clients' unlabeled data is thus of paramount practical importance. In this paper, we propose a new FL algorithm, called FedSEAL, to solve this Semi-Supervised Federated Learning (SSFL) problem. Our algorithm utilizes self-ensemble learning and complementary negative learning to enhance both the accuracy and the efficiency of clients' unsupervised learning on unlabeled data, and orchestrates the model training on both the server side and the clients' side. Our experimental results on Fashion-MNIST and CIFAR10 datasets in the SSFL setting validate the effectiveness of our method, which outperforms the state-of-the-art SSFL methods by a large margin.