Department of Physics, West Cambridge

The role is funded by the "Unleash machine learning for materials discovery" programme. The project will develop a new form of federated machine learning that, merges separate machine learning models, before applying the methodology to collaborator's data. The research will be hosted in Dr Conduit's group within the Department of Physics which has a track record of developing machine learning tools, designing experimentally verified materials leading to six patents and many peer-reviewed papers. The approach has been commercialized by Intellegens as the product Alchemite Analytics for use in materials, industrial chemicals and pharmaceuticals. This project will continue the development of new approaches to federated learning, that offers significant practical advantages including the ability to merge not only machine learning models with different architectures, but also computational methods and analytical formulae. The machine learning models can be trained in parallel, readily withdrawn from the merged model and value ascribed to each model's contribution.

The applicant will be required to work with experimental collaborators across multiple science disciplines to demonstrate the method's benefits and develop case studies leading to peer-reviewed publications. The studies may highlight limitations of federated learning approaches, which will motivate the post holder to pursue and test further algorithmic developments.

https://www.jobs.cam.ac.uk/job/34500/