Finding fast and accurate algorithms for the modern world
A new collaboration designed to produce new algorithms for use in big data sets will be led by Dr Isolde Adler of the School of Computing.
Fast and Accurate! Foundations and Guarantees for Algorithms and AI (FGAA) is a collaboration funded by the White Rose University Consortium, a partnership between the Russell Group universities of Leeds, Sheffield and York.
The project aims to provide new algorithms for big data sets and networks that are both extremely fast and come with accuracy and performance guarantees.
As data and the ways in which we use it become increasingly complex, there is an urgent need for algorithms (computing instructions) that can tackle more and more resource-intensive computational problems.
Even seemingly simple examples like the famous Travelling Salesperson problem, which seeks to calculate the shortest possible circular route between a series of points, can be very difficult for today’s algorithms to compute. In extreme cases, even the fastest Travelling Salesman algorithm can take years to calculate an optimal route.
With algorithms now driving everything from cyber-security to autonomous vehicles, calculating high-quality routes within milliseconds can save time, money, energy and even lives in an emergency. Crucially, these algorithms must also come with guarantees regarding the quality of the computed answer, the running time and energy consumption, as well as reliability, e.g. of control algorithms.
The FGAA project will bring together researchers from diverse backgrounds to address these problems. It proposes a novel combination of classical and recent methods (including sublinear time algorithms, local, distributed, randomised, parametrised, machine learning) from the areas of theoretical computer science (with mathematical proofs providing guarantees), discrete algorithms, logic and artificial intelligence.
Dr Isolde Adler will lead the project, together with Dr Sebastian Ordyniak of the University of Sheffield and Dr James Cussens of the University of York. Dr Haiko Muller and Dr Felix Salfelder (Leeds), Dr Pan Peng (Sheffield) and Dr Peter Nightingale (York) are also associated with the project.
