Enregistré dans:
Détails bibliographiques
Auteurs principaux: Park, Byoungjun, de Gusmão, Pedro Porto Buarque, Ji, Dongjin, Kim, Minhoe
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:https://arxiv.org/abs/2511.11778
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866911275870584832
author Park, Byoungjun
de Gusmão, Pedro Porto Buarque
Ji, Dongjin
Kim, Minhoe
author_facet Park, Byoungjun
de Gusmão, Pedro Porto Buarque
Ji, Dongjin
Kim, Minhoe
contents Federated learning is a promising paradigm that utilizes distributed client resources while preserving data privacy. Most existing FL approaches assume clients possess labeled data, however, in real-world scenarios, client-side labels are often unavailable. Semi-supervised Federated learning, where only the server holds labeled data, addresses this issue. However, it experiences significant performance degradation as the number of labeled data decreases. To tackle this problem, we propose \textit{CATCHFed}, which introduces client-aware adaptive thresholds considering class difficulty, hybrid thresholds to enhance pseudo-label quality, and utilizes unpseudo-labeled data for consistency regularization. Extensive experiments across various datasets and configurations demonstrate that CATCHFed effectively leverages unlabeled client data, achieving superior performance even in extremely limited-label settings.
format Preprint
id arxiv_https___arxiv_org_abs_2511_11778
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle CATCHFed: Efficient Unlabeled Data Utilization for Semi-Supervised Federated Learning in Limited Labels Environments
Park, Byoungjun
de Gusmão, Pedro Porto Buarque
Ji, Dongjin
Kim, Minhoe
Machine Learning
68T05
Federated learning is a promising paradigm that utilizes distributed client resources while preserving data privacy. Most existing FL approaches assume clients possess labeled data, however, in real-world scenarios, client-side labels are often unavailable. Semi-supervised Federated learning, where only the server holds labeled data, addresses this issue. However, it experiences significant performance degradation as the number of labeled data decreases. To tackle this problem, we propose \textit{CATCHFed}, which introduces client-aware adaptive thresholds considering class difficulty, hybrid thresholds to enhance pseudo-label quality, and utilizes unpseudo-labeled data for consistency regularization. Extensive experiments across various datasets and configurations demonstrate that CATCHFed effectively leverages unlabeled client data, achieving superior performance even in extremely limited-label settings.
title CATCHFed: Efficient Unlabeled Data Utilization for Semi-Supervised Federated Learning in Limited Labels Environments
topic Machine Learning
68T05
url https://arxiv.org/abs/2511.11778