Advanced Machine Learning

Link to course: 7MRI0010

Term: Semester 2

Lectures: Will be made available open-source soon.


To educate students with regards to novel artificial intelligence algorithms for the analysis and predictive modelling of multiple types of healthcare data such as medical images, genetics, clinical/epidemiological variables, and free text. The module provides the theoretical understanding and practical implementation of models that are at the core of advanced analysis and actionability of medical data.

The module utilizes a mixture of lectures, required reading with associated seminars, tutorials, and labs. The lab sessions and tutorials will be used to expose students to the real-world need and limitations of advanced models. From applying simple feed-forwards convolutional neural networks in a supervised setting, to training uncertainty-aware complex multi-task networks from multiple sources of data. The core material will include, but are not limited to the following:

  • New representation learning architectures and loss functions
  • Adversarial learning in a medical setup
  • Attention and auxiliary tasks as a form of regularisation.
  • Model interpretability and introspection
  • Uncertainty estimation using Bayesian deep learning
  • Reinforcement learning in medicine
  • Sequence and predictive modelling
  • Natural language processing and language encoding
  • Multi-modal predictive models
  • Validation of AI models in healthcare

Students will work on the assignments both individually and in groups, but reports will be submitted independently. This guarantees that each student will learn the fundamental concepts as will be assessed in the report, but they will be able to collaborate towards producing a working implementation of complex algorithms. Tutor support will be organized with associated workshops. This module will run throughout semester 2 and links well with the Machine Learning for Biomedical Applications module of semester 1