Agenda
09:00 – 09:20 Registration
09:20 – 09:30 Opening Remarks
Professor Sophia Yaliraki, Imperial
09:30 – 10:30 AI Deep Dive Session | Two Focus Talks
Chair: Sophia Yaliraki
Tolga Birdal, Imperial: Topological Deep Learning: A New Hope for Molecular Modelling and 3D Vision
Paula Cordero Encinar, Imperial: Sampling by Averaging: A Multiscale Approach to Score Estimation
10:30 – 11:00 Coffee Break
11:00 – 12:00 Flash Talk Session | Five 10-Minute Flash Talks
Chair: Ben Moseley
Kevin Han Huang, University College London: Diagonal Symmetrization of Neural Network Solvers for the Many-Electron Schrödinger Equation
Cristiana Diaconu, University of Cambridge: Gridded Transformer Neural Processes for Spatio-Temporal Data
Hugh Dance, University College London: Efficiently Vectorized MCMC on Modern Accelerators
Yedi Zhang, University College London: Training Dynamics of In-Context Learning in Linear Attention
Lisa Alazraki, Imperial: Reverse Engineering Human Preferences with Reinforcement Learning
12:00 – 13:00 Lunch and Poster Session
13:00 – 14:00 AI Deep Dive Session | Two Focus Talks
Chair: Laura M. Helleckes
Naoki Kiyohara, Imperial: Neural Stochastic Flows: Solver-Free Modelling and Inference for SDE Solutions
Lauri Laatu, Imperial: Sub-Microsecond Transformers for Jet Tagging on FPGAs
14:00 – 14:30 Coffee Break
14:30 – 15:30 Flash Talk Session | Five 10-Minute Flash Talks
Chair: Daniel Platt
Marta Aparicio Rodriguez, Imperial: Concept Reachability in Diffusion Models
Guiomar Pescador Barrios, Imperial: Adjusting Model Size in Continual Gaussian Processes: How Big is Big Enough?
Jan Ole Ernst, University of Oxford: Reinforcement Learning for Quantum Control Under Physical Constraints
Aya Kayal, University College London: Bayesian Optimization from Human Feedback: Near-Optimal Regret Bounds
Krzysztof Kacprzyk, University of Cambridge: Skip the Equations: Learning Behavior of Personalized Dynamical Systems Directly from Data
15:30 – 16:00 Coffee Break
16:00 – 17:00 AI Deep Dive Session | Three Focus Talks
Chair: Nick Jones
Eleonora Giunchiglia, Imperial: Right for the Right Reasons: Avoiding Reasoning Shortcuts via Prototype-Augmented Neurosymbolic AI
Peter Ochieng, University of Cambridge: Diversity Is All You Need for Contrastive Learning: Spectral Bounds on Gradient Magnitudes
Georgia Channing, University of Oxford: MAD-Sherlock: Multi-Agent Debate for Visual Misinformation Detection
17:00 – 17:10 Closing Remarks
Professor Nick Jones, Director of I-X Centre for AI in Science, Imperial
