Module

Module texts ( and further reading)

Mitchell,T.,  Machine Learning, Mcgraw-Hill International edition

Submission:

Written coursework should be submitted by following the standard CMP practice. Students are advised to refer to the Guidelines and Hints on Written Work in CMP.

Deadlines:

If coursework is handed in after the deadline day or an agreed extension:
 

Saturdays and Sundays will NOT be taken into account for the purposes of calculation of marks deducted.

All extension requests will be managed through the LTS Hub. A request for an extension to a deadline for the submission of work for assessment should be submitted by the student to the appropriate Learning and Teaching Service Hub, prior to the deadline, on a University Extension Request Form accompanied by appropriate evidence. Extension requests will be considered by the appropriate Learning and Teaching Service Manager in those instances where (a) acceptable extenuating circumstances exist and (b) the request is submitted before the deadline. All other cases will be considered by a Coursework Coordinator in CMP.

For more details, including how to apply for an extension due to extenuating circumstances download Submission for Work Assessment (PDF, 39KB)
 

Plagiarism:

Plagiarism is the copying or close paraphrasing of published or unpublished work, including    the work of another student; without due acknowledgement. Plagiarism is regarded a serious offence by the University, and all cases will be investigated. Possible consequences of plagiarism include deduction of marks and disciplinary action, as detailed by UEA's Policy on Plagiarism and Collusion.

Module specific:

• To understand the nature of classification, clustering and reinforcement learning problems and to have an overview of the areas of business and science in which they may occur.
• To comprehend the workings of the classification, clustering and reinforcement algorithms covered in the course and the motivation for why the work in the way they do
• To be able to execute by hand simple versions of these algorithms on toy problems
• To be able to implement basic versions of these algorithms that can be applied to real world problems
• To be able to use a variety of tools with fully implementations of these algorithms
• To be able to usefully compare the performance of these algorithms and to grasp which algorithms work best for which type of problem
• To be able to present these results in a logical, scientific way to the owner of the problem.

Transferable skills:

• To gain further experience in understanding algorithms
• To improve programming skills
• To learn how to understand and utilize complex existing code.
• To understand how to logically structure a report describing a scientific approach to problem solving

On completion of this module students should be able to:

• Formulate problems as classification or clustering problems and apply a wide range of algorithms to the task
• Write a coherent and accurate summary of how the different algorithms perform on a problem and suggest knowledge that may be inferred from the analysis
• Propose potentially interesting ways the algorithms covered could be extended as a research project

Total hours: 45 (approx.)

Lectures:  20-30 hours

1. Introduction: Basic Principles (simple univariate classifiers) Naive Bayes
2. Nearest Neighbour Classifiers
3. Decision Trees
4. Linear Classifiers
5. Artificial Neural Networks
6. Support Vector Machines, Performance Evaluation
7. Bayesian Networks
8. Unsupervised learning 1: clustering and PCA
9. Unsupervised learning 2: reinforcement learning
10. Learning Classifier Systems for supervised and unsupervised learning

Workshops: 5-10 hours

Laboratory work: 10-20 hours