2024 - 2025 Course Highlights
Dissertations
Computer Vision and Robotics:
PhysMo: Physics-based Animal Motion Reconstruction from Videos
Pre-trained Visual Representations Generalise Where it Matters in Model-Based Reinforcement Learning
Machine Learning:
Are Variants of the CRPS Sufficient for Generative Modelling?
Audio Interpretable Transformers
Continual Decision Making in Recurrent Neural Networks
Evaluating Bias Awareness in Large Language Models
Incremental Learning in Transformer Neural Processes
Interpretable Language Transformers via Audio Embeddings
Learning Non-Parametric Structural Decompositions in Gaussian Processes
Multiscale Latent-Space Models for Efficient Medium Range Probabilistic Weather Forecasting
Rotary Positional Encodings for Graph Transformers
Stochastic Interpolants in Hilbert Spaces
Streamable Diffusion Model for Speech Separation
Transferable Neural Transports for Parallel Tempering
Speech and Language Processing:
Automatic Speech Recognition for Children's Speech
Bias Dynamics in BabyLMs: Towards a Compact Sandbox for Democratising Pre-Training Debiasing
Enhancing Knowledge Distillation between Large Language Models with Vocabulary Mismatch
Human-Computer Interaction:
Developing a Vision-Language-Action Model to Drive Contextual User Interface Adaptation