DERI Seminar with Shaheer Ullah Saeed
When: Thursday, October 15, 2026, 11:00 AM - 12:00 PM
Where: on Teams, on Teams
Speaker: Dr Shaheer Ullah Saeed, Lecturer in Computational Biomedical Engineering at QM
Title: Teaching machines to reason by giving them time to think
Abstract: Reasoning is a hallmark of human intelligence, enabling adaptive decision-making in unfamiliar situations. Machine intelligence, however, remains largely bound to training data, with limited ability to refine solutions at inference. Advances in machine reasoning address this by trading additional thinking time for improved performance, but largely focus on language, mathematics and board games. In this talk, we will present a framework for reasoning in vision, inspired by human fast and slow thinking. It combines rapid solution generation and verification with iterative refinement through self-play reinforcement learning. Across computer-vision benchmarks and cancer localisation in five organs, extended thinking time yields performance surpassing supervised models, foundation models and human experts, even with limited labelled data.
Bio: Shaheer U. Saeed is a Lecturer in Computational Biomedical Engineering, and Cancer Research UK Pathway Award Fellow, at Queen Mary University of London. His research focuses on developing data-efficient learning algorithms for cancer detection, with particular emphasis on enabling AI systems to learn and reason from limited and imperfect medical imaging data. His work combines reinforcement learning, meta-learning and machine reasoning to improve the reliability and clinical applicability of medical AI.
________________________________________________________________________________
Microsoft Teams meeting
Join: https://teams.microsoft.com/meet/3978780609865781?p=QUwVZ16GnBTGNuDJ6O
Meeting ID: 397 878 060 986 578 1
Passcode: 4nb3Fi3r
For organizers: Meeting options
________________________________________________________________________________