The University of Sheffield
Department of Computer Science

COM6509 Machine Learning and Adaptive Intelligence

Summary The module is about core technologies underpinning modern artificial intelligence. The module will introduce statistical machine learning and probabilistic modelling and their application to describing real-world phenomena. The module will give students a grounding in modern state of the art algorithms that allow modern computer systems to learn from data.
Session Autumn 2019/20
Credits 15
Assessment

40% Formal examination
60% Coursework, consisting of assignments and quizzes.

Lecturer(s) Dr Mauricio Alvarez & Dr Haiping Lu
Resources
Aims This unit aims to provide a deep understanding of the fundamental technologies underlying modern artificial intelligence. In particular, it will provide a foundational understanding of
  • data science,
  • probability in artificial intelligence.
Objectives By the end of the unit, a candidate will be able to demonstrate the ability to
  • understand of probability theory and how it relates to uncertainty in modern artificial intelligence,
  • understand when and how to implement the appropriate learning paradigm for a given application,
  • have a deep understanding (including how to implement) of a range of supervised learning algorithms. Potential examples include: linear regression, linear classification, naive Bayesian classification, principal component analysis, k-means clustering, decision trees.
  • have a broad overview and basis for understanding more complex technologies, such as: the support vector machine, kernel methods, probabilistic graphical models, E-M algorithms, factor analysis, nonparametric Bayesian methods.
Content
  • Probability theory
  • Objective function
  • Linear regression
  • Basis functions
  • Generalisation
  • Bayesian regression
  • Unsupervised learning
  • Probabilistic Classification
  • Logistic Regression and Generalised Linear Models
  • Other topics
Teaching Method There will be 10 hours of formal lectures and 20 hours of lab classes. These classes may involve computer-based problem solving, examples and MCQ. The lectures and example classes will run concurrently, covering similar topics, with the theory being laid out in the lectures, and application of the theory taking place in example classes.
Feedback Students will receive feedback on their assignments and quizzes through MOLE and in the Lab Sessions.
Recommended Reading
  • S. Rogers and M. Girolami, A First Course in Machine Learning
  • C. Bishop, Pattern Recognition and Machine Learning
  • D. Barber, Bayesian Reasoning and Machine Learning
  • R. Duda, P. Hart and G. Stork, Pattern Classification