Contact person
Olof Mogren
Senior Researcher
Contact OlofFederated learning
Federated learning take advantage of data that is distributed across a number of clients. Learning takes place locally on the clients, and with the help of a central server these clients can collaborate and achieve an even better generalization. This provides privacy benefits, as well as distributed compute.
Many applications require the data to stay where it is. This may be for technical reasons such as limited network capacity or computing power, or for legal reasons.
In these cases, distributed solutions may enable advanced analysis of data that would otherwise be impossible. Federated learning also offers the opportunity to take advantage of many compute nodes that work together to learn a problem. This can provide strengths in computing power and energy efficiency.
RISE has strong expertise and actively invests in distributed learning. We develop techniques for efficiency based on distributed computing power, but also to be able to provide guarantees or quantifications of privacy in these settings.
We currently have collaborations where we apply federated strategies for the telecom industry, as well as in medical applications.
Related publications:
Listo Zec, E., Mogren, O., Martinsson, J., Sütfeld, L.R., Gillblad, D. (2020) Federated learning using a mixture of experts. https://arxiv.org/abs/2010.02056
