A team of UK scientists from the universities of Bath, Cambridge and UCL aims to make Deep Learning (DL) more accountable and transparent by better understanding the decision making process behind the algorithms.
The team of mathematicians, statisticians and image processing experts has been awarded a five-year Programme Grant worth £3.5M by the EPSRC on ‘The Mathematics of Deep Learning’ (Maths4DL), starting in January 2022.
Machine learning, in particular Deep Learning (DL) based on ‘neural networks’, is one of the fastest growing areas of modern science and technology, which has potentially an enormous and transformative impact on all areas of our lives.
The applications of DL embrace many disciplines such as biomedical sciences, computer vision, the physical sciences, the social sciences, speech recognition, gaming, music and finance.
However, alongside this explosive growth has been a concern about the lack of understanding behind DL and the way that DL based algorithms make their decisions. This leads to a lack of trustworthiness in the use of some of these algorithms.
A reason for this is that the huge successes of Deep Learning are not all well understood, the results are sometimes mysterious, and there is often a lack of a clear link between the data training DL algorithms, and the decisions made by those algorithms.
This Programme Grant aims to put DL onto a firm mathematical grounding, and will combine theory, modelling, data and computation to help unlock the next generation of deep learning.
Professor Chris Budd, from the University of Bath's Department of Mathematical Sciences and Director of Knowledge Exchange at the University’s Institute for Mathematical Innovation (IMI), will lead the team.
He said: “Huge advances in machine learning mean that DL algorithms could be used for a wide range of applications, such as helping diagnose illness, driving cars and even making legal judgements. The possible applications in the future are almost unlimited.
“We urgently need the opportunity to improve our understanding of machine learning. This programme grant aims to rise to this challenge, and, by doing so, to unlock the future potential of artificial intelligence.”
The research work in the grant will comprise an interlocked set of work packages aimed to address both the theoretical development of DL (so that it becomes explainable) and the algorithmic development (so that it becomes trustworthy).
These will then be linked to the development of DL in a number of key application areas, linked to and supported by industry, including medical image processing, partial differential equations and environmental problems.
The team will work closely with industrial partners, colleagues at Bath’s Centre for Mathematics and Algorithms for Data, the public, and policy makers.