We have an increasingly good foundation of data collection and analysis at SIB, and we will continue to build on that foundation in the coming year. A brief snapshot of some of our forthcoming work is below:
Methodology – Etic Lab
We don’t just want to be a sector lead on data, but continue to partner with our peers to help them use their data more effectively. Etic Lab is a digital research and design consultancy that helps facilitate collaboration in the face of unhelpful circumstances, particularly with regard to the use and sharing of data. Their hope in attempting this work is to open up the use of data analytics, machine learning and other similar techniques to organisations who would otherwise not have the opportunity to employ them, whether because they lack sufficient data or other resources to attempt such projects on their own, due to concerns relating to the privacy and security of private or proprietary information, or the myriad other factors which can stand in the way of the strategic use of data assets.
By setting up a partnership with Etic Lab, we are hoping to facilitate the work of SIB’s Data Lab through the use of federated learning and other techniques.
Federated Learning allows for the creation of machine learning models in situations where - whether for reasons of privacy, competitive advantage or scale - it is not feasible to collect and hold data in a single location. It also reduces the requirement for data to be standardised across contributing datasets. In this way, it opens up the use of this technology to a range of organisations who would not previously have been able to access it.
A group of partners contribute data to train a shared prediction model. All contributed data remains securely on the computer systems of the participants, and is never exposed or collected in a central database. Instead, the learnings derived from each set are extracted and aggregated to produce a single model based on the totality of all contributed data. The resultant model is the shared property of all partners, who are entitled to use it according to the terms of a pre-established agreement. It has relatively modest hardware requirements, and can be easily extended or updated with the addition of new data. Federated learning therefore responds to needs identified through the Data Lab and its partners so far, where some partners have expressed privacy concerns, or do simply not have available resource to manage data input.