Date: June 30, 2021
Time: 4 pm BST / 11 am EDT / 8 am PDT
Federated learning is a new machine-learning paradigm where multiple partners can collaborate on complex research questions without centralizing or sharing data outside of their organizations.
This ‘collaborative machine learning’ approach enables data science teams to work on larger and more diverse datasets, previously inaccessible, boosting the predictive power of machine learning algorithms and enhancing AI capabilities. By overcoming privacy and confidentiality concerns, companies can build partnerships and consortia and retain their competitive edge.
For example, the MELLODDY consortium pioneers federated learning-based drug discovery across 10 pharma companies benefiting from the collective insights of the world’s largest cheminformatics data network where each participant retains full confidentiality and governance over their molecular libraries.
Federated learning in healthcare can also facilitate knowledge transfer between medical researchers and data scientists, bridging the gap between AI and clinical care. The HealthChain project is a successful demonstration that an algorithm can be trained on siloed histology images, distributed across different hospitals, to predict treatment responses in breast cancer. Together with clinical, research, and technology partners, we demonstrated improved robustness and performance of the technology over locally trained algorithms.
With the platform deployed and used reliably in a production environment, the stage is set for further collaborative research projects and eventually clinical applications in cancer, heart failure, and other therapeutic areas.
- Victor Dillard, Commercial Operations Director, Owkin
- Hugo Ceulemans, Scientific Director Discovery Data Sciences, Janssen
- Dr. Guillaume Bataillon, Pathologist, IUCT Oncopole