HLTH 667MGenerative AI Core Techniques, Architectures and Algorithms for Data Representation and Generation

This course provides an introduction to the computational and mathematical foundations of Generative Artificial Intelligence (GenAI), with an emphasis on the models and algorithms that underpin modern generative systems. Students will learn the principles behind core generative model families, including autoencoders, variational autoencoders (VAEs), generative adversarial networks (GANs), autoregressive models, normalizing flows, diffusion models, and transformer-based large language models. Topics include latent representations, probabilistic modelling, attention mechanisms, optimization strategies, training objectives, and challenges such as mode collapse, stability, and scalability. Tutorial sessions will guide students through the implementation of selected generative architectures in Python and PyTorch, exploring training dynamics, evaluating generated outputs, and understanding model behaviour through visualization and diagnostics.