ENGINEERING TRIPOS PART IIB - 2009/2010

Module 4F13 - Machine Learning

AIMS

Machine learning (ML) is an interdisciplinary field focusing on both the mathematical foundations and practical applications of systems that learn, reason and act. ML builds on ideas from Engineering, Computer Science, Statistics and Cognitive Science, and has many practical applications to problems such as speech recognition, computer vision, robotics, information retrieval, computational finance, and bioinformatics. The aim of this module is to introduce students to basic concepts in machine learning, forcusing on statistical methods, unsupervised learning and reinforcement learning. One lecture in the middle of this module will be devoted to example applications of machine learning.

SYLLABUS

1. Introduction to Machine Learning (1L)

• review of probabilistic models, relation to coding terminology
• Bayes rule, maximum likelihood, MAP
• supervised. insupervised and reinforcement learning

2. Unsupervised Learning (2L)

• factor analysis, PCA, Independent Components Analysis (ICA), the EM algorithm

3. Bayesian Networks (3L)

• conditional independence, Belief Propagation and relation to HMMs, Kalman filters

4. Applications of Machine Learning (1L)

• gene microarrays, information retrieval, speech recognition, character recognition, collaborative filtering, medical diagnosis, financial prediction, and text input (Dasher)

5. Monte Carlo Methods (2L)

• simple Monte Carlo, rejection sampling, importance sampling, Metropolis, Gibbs sampling, hybrid Monte Carlo

6. Variational Approximations (1L)

• KL divergences, mean field, expection propagation

7. Model Comparison (1L)

• Bayes factors, Occam's razor, BIC Laplace approximations

8. Reinforcement Learning, Decision Making and MDPs (3L)

• value functions, value iteration, policy iteration
• Bellman equations, Q-learning, Bayesian decision theory

Lectures will be supported by MATLAB demonstrations.

OBJECTIVES

On completion of the module students should:

• Have a good understanding of basic concepts in statistical machine learning
• Be able to apply basic ML methods to practical problems

REFERENCES

Please see the Booklist for Group F Courses for references for this module.

Last updated: September 2009