Abstract
Feature selection is essential for interpretable and robust learning from high-dimensional, low-sample-size data. This setting is common in biomedical and clinical studies, where the number of variables is large, cohorts are small, and data sharing across institutions is often restricted. Federated learning (FL) enables collaborative model training without sharing raw data. However, feature selection remains challenging in federated settings. Locally sparse models often identify site-specific predictors, while feature-election methods aggregate locally selected masks without jointly optimizing a sparse predictive model. Here, we introduce LAFFS, a horizontal federated embedded feature-selection framework that integrates LASSO regularization into collaborative model training. LAFFS jointly trains local and global models while imposing sparsity on the aggregated global representation. This strategy encourages participating sites to recover a consistent set of globally relevant features. We evaluated LAFFS against traditional machine-learning methods, federated feature-selection approaches, and feature-election methods using matched site partitions and validation protocols. LAFFS achieved competitive predictive performance in high-dimensional tasks. Our results demonstrate that feature selection in federated settings can be formulated as a native federated optimization problem.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/DeCaF_015.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=m43wrB97nv
BibTex
@InProceedings{ShaGau_LAFFS_MICCAISAT2026,
author = { Sharma, Gaurang AND Moradi, Elaheh AND Pajula, Juha AND Hilvo, Mika AND Tohka, Jussi},
title = { { LAFFS: LASSO Assisted Federated Feature Selection } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17252},
month = {pending},
page = {pending}
}