Survival Model Optimization via Federated Learning: A Study Combining Simulations and Experiments
Federated Learning is an emerging, powerful approach that allows training an artificial intelligence model in distributed setting. Two survival models, Cox and DeepSurv, have been trained in a federated setting, exploiting both code simulations and real experiments on the new platform, developed by the GenoMed4All consortium. Different scenarios have been tested by splitting a Myelodysplastic Syndrome dataset into three nodes and performing feature removal. A significant gain in model performance has been observed due to federated aggregation.
Authors
- Cesare Rollo (ORCID: https://orcid.org/0000-0001-6093-1454)
- Santiago Zazo (ORCID: https://orcid.org/0000-0001-9073-7927)
- Tiziana Sanavia (ORCID: https://orcid.org/0000-0003-3288-0631)
- Saverio D’Amico (ORCID: https://orcid.org/0000-0003-1877-8497)
- Enrico Giampieri (ORCID: https://orcid.org/0000-0003-2269-2338)
- Gastone Castellani (ORCID: https://orcid.org/0000-0003-4892-925X)
- Patricia A. Apellániz
- Claudia Sala (ORCID: https://orcid.org/0000-0002-4889-1047)
- Federico Álvarez García
- Juan Parras (ORCID: https://orcid.org/0000-0002-7028-3179)
- Piero Fariselli (ORCID: https://orcid.org/0000-0003-1811-4762)
- Luciana Carota (ORCID: https://orcid.org/0000-0002-9217-4600)
- Nono S. C. Merleau (ORCID: https://orcid.org/0000-0003-3150-6250)
- Matteo Giovanni Della Porta (ORCID: https://orcid.org/0000-0002-6915-5970)
- Gianluca Asti (ORCID: https://orcid.org/0009-0006-9078-6482)
- Davide Piscia
- Francesco Casadei (ORCID: https://orcid.org/0009-0006-0789-546X)
Institutions
- Humanitas University (IT)
- Max Planck Institute for Mathematics in the Sciences (DE)
- Azienda USL di Bologna (IT)
- Max Planck Institute for Mathematics (DE)
- Istituto delle Scienze Neurologiche di Bologna (IT)
- Istituti di Ricovero e Cura a Carattere Scientifico (IT)
- University of Bologna (IT)
Publication Details
- Published
- 2024-12-15
- DOI
- https://doi.org/10.1109/bigdata62323.2024.10825368
- Primary Topic
- Privacy-Preserving Technologies in Data
- Type
- article
- Field-Weighted Citation Impact
- 0.00