Assured Systems (UK) Ltd are a leading technology company offering high quality and innovative computer solutions to the embedded, industrial, and digital-out-of-home market sectors. Please visit Assured System’s webpage for further details: www.assured-systems.com/uk.
Machine Learning (ML) involves training computers to learn from data in order to act without human intervention. Traditionally, this is performed in the ‘cloud’ which can be a problem when sensitive data is involved. Researchers at Google pioneered a method called federated learning, which allows many computing devices to collaboratively train a machine learning model on their local data. Therefore, the data never needs to leave the user’s device.
This project is applying federated learning techniques to Internet of Things (IoT) applications. The results could be used for forecasting energy demand in smart grids while protecting household’s raw smart meter data. The results could also be used for energy analytics, such as specific appliance demand profiles, without revealing which households participated in the dataset.
The project could have many applications to the SEND project. One example would be providing SEND with a way to see how energy demand is likely to fluctuate within residential homes on campus in order optimise the network, while maintaining the privacy of the data from the individual households.
Project progress (June 2019):
- Submission of literature review (Feb 2019)
- Presentations and poster presentations at various Keele internal conferences and symposia (Mar 2019 – May 2019)
- Submission of review paper to ACM Transactions on Intelligent Systems & Technology (TIST) journal (Apr 2019)
- Re-implementation of state-of-the-art federated learning algorithm to recreate the results of McMahan et al 2017 (Ongoing)
Business Contact(s): James Priest
Keele Graduate Researcher: Chris Briggs