This module introduces modern deep learning methods for analysing structured data, images, and sequences, with a strong balance between theory and practice. Core deep learning concepts are covered, including neural network architectures, training strategies, and optimisation, alongside practical considerations for building effective models.
Image analysis is explored in depth, focusing on representation learning and computer vision applications. The module also addresses language modelling for sequential data, reinforcement learning for decision-making through interaction, and meta-learning approaches that enable models to learn efficiently from limited data. Advanced topics include physics-based and physics-informed deep learning, probabilistic neural networks for uncertainty-aware modelling, and diffusion models for generative data and image analysis.
A defining feature of the module is its hands-on approach: regular computer lab sessions provide experience with modern deep learning frameworks, while coursework emphasises practical implementation, experimentation on real datasets, and critical evaluation of model performance
