Boosting Algorithm Performance with Imperfect Advice. This project aims to develop a suite of new methods to design algorithms that perform best with AI-derived advice, even when imperfect. Recent adv

Description

Boosting Algorithm Performance with Imperfect Advice. This project aims to develop a suite of new methods to design algorithms that perform best with AI-derived advice, even when imperfect. Recent advances in AI show its power and compel its application. This project is a significant and timely contribution to knowledge, with principled methods of integrating AI into algorithms, without sacrificing safety or robustness. Expected outcomes include practical algorithms with provable guarantees that leverage advice to achieve better solution quality and memory utilisation, especially when data is streamed. These new algorithms will provide significant benefits in effectiveness and efficiency gains, and robustness guarantees, for real-world problems arising in transport, energy, and cybersecurity.. Scheme: Discovery Projects. Field: 4613 - Theory of Computation. Lead: Prof Anthony Wirth