Machine Learning / Extended – 2015/16

Syllabus pages: Machine Learning, Machine Learning Extended – these pages include the recommended textbooks.

CA marks: Bonus 1 Bonus 2 ClassTest+Take Home Test. The Type 2 Assignment marks are in Canvas along with feedback comments where applicable.

Math refreshers:

Linear algebra refresher

Probability theory refresher

Lecture handouts:

I.     Introduction. Admin & Structure of the module; Induction slides
Check out: Tom Mitchell: The discipline of Machine Learning & a recent short video

II.   Supervised learning methods

o    Methods of classification: generative, discriminative; global, local; probabilistic, non-probabilistic

1.   Bayes rule. Intro to generative classification. + Exercise Sheet (non-assessed) Solutions

2.   Generative classifiers: The Gaussian classifier + Example

Type 2 Assignment 1 + Code + gauss.m

Exercise Class Notes

3.       K-Nearest Neighbours methods Worksheet (non-assessed) Solutions

4.   Discriminative classifiers: Intro to Support Vector Machines + Worksheet (non-assessed) Solutions

Suite of exercises preparatory for the class test. Solutions – guest exercise class by Momodou Sanyang

The Class Test was on Friday 13 November. Open book. [15%] Test questions + Model solutions

5.   Evaluating classifiers

Type 2 Assignment 2 + data + code

III.  Unsupervised learning methods

o    Gaussian mixtures and the EM algorithm – guest lecture by Alastair Turl

o    Clustering by k-means

Type 2 Assignment 3 + code&data

IV. Intro to Statistical Learning Theory: The PAC model of learning in finite hypothesis classes + proof

Take home test Test questions [5%]

V.  Ensemble methods: Boosting

Type 2 Assignment 4 + code

Learn more: Top 10 Algorithms for Data Mining

  1. Topics in learning from high dimensional data and large scale learning + proof

 

Additional resources

A past exam paper: Main Resit

Example exam paper from 2012 (obsolete, since ML was a 2nd year module and the content is now different)

Andrew Ng on Coursera

The `Matrix Cookbook’ – useful matrix and vector manipulation formulae

Notes from my talk at SoCS Training for Research: Machine Learning Concepts & Techniques

MatLab tutorials:
http://www.cyclismo.org/tutorial/matlab/
http://users.rowan.edu/~shreek/networks1/matlabintro.html
http://www.mathworks.com/access/helpdesk/help/pdf_doc/matlab/getstart.pdf

Previous years Machine Learning page – for the curious – now obsolete.