This course introduces the area of machine learning from a statistical perspective. Understanding modern machine learning algorithms is of huge importance, and requires a deep understanding of statistical theory and algorithms. The course will present the necessary mathematical tools required which are divided primarily into variational and (mostly) statistical methodologies. Examples of this include optimisation, which is based on the notion of convexity, where the students will be introduced to stochastic gradient methods. For the latter, the main focus of the statistical component is on the discussion of Markov chains and how one can use these entities to sample from probability distributions. Once the students become aware of this they will then discuss topics in statistical machine learning such as neural networks, deep learning, statistical learning theory and empirical risk minimization. Overall the course provides a solid introduction of statistics that underpins machine learning, from an analytical, methodological and application viewpoint.

40% Coursework

60% Examination (Duration: 2 hours, at the end of the semester)

For a student to pass the course, at least 30% of the maximum mark for the examination must be obtained.