Skip to main content
Advanced Technical IVT

federated learning

Machine learning approach that enables model training across distributed datasets without centralizing patient data, important for multi-institutional oncology research.

Full Definition

Federated learning is a machine learning paradigm that enables the training of algorithms across decentralized data sources without requiring data to leave its original location. In oncology informatics, this approach is particularly valuable for multi-institutional research where patient privacy regulations and institutional policies prevent data sharing. Each participating site trains a local model on their data, and only the model parameters or updates are shared with a central coordinator. This enables the development of robust predictive models using larger, more diverse datasets while maintaining patient privacy and data security. Applications include predicting treatment responses, identifying optimal treatment protocols, and developing diagnostic algorithms.

Usage

Usage note: Emerging technology in healthcare AI; may require explanation of privacy-preserving aspects.

In Context

  • "The cancer research consortium used federated learning to develop a predictive model without sharing sensitive patient data between institutions." — Research collaboration agreement
  • "Federated learning enabled the training of imaging AI models across multiple cancer centers while maintaining HIPAA compliance." — Technical implementation report

Also known as

collaborative learning distributed machine learning

Contrasted with

centralized learning

Don't confuse with

distributed computing cloud-based analytics

Editors from these organisations have used our services since 1998

Reuters BBC Oxford University Press Penguin Random House Springer Microsoft Suncor Energy United Nations Fisher Investments IBM The Home Depot KODAK CHEVRON