Overview

Machine Learning is a sub-field of both statistics and artificial intelligence. With the availability of massive amounts of data and computational power, what was once a fairly obscure affair has now acquired a certain notoriety (think of “deep learning”). The huge number of algorithms used in machine learning turn out to be almost all increasingly subtle variations of four or five basic ideas, and these are the subject of our course. Each of these ideas will be introduced with an example, we will then examine its simplest mathematical formulations and try to understand when it works, and why.

A word of warning: this is not a “hands-on” introduction to the various software packages that are used for machine learning.

Knowledge of programming is not necessary, beyond the ability to understand and formulate algorithms. I will occasionally show an algorithm in action, but will not give detailed instructions on installing or using software, and use of a computer is not essential to following this course.

Programme details

Week 1: Introduction, Concept learning

Week 2: Bayesian Learning 1

Week 3: Bayesian Learning 2

Week 4: Unsupervised learning

Week 5: Neural Networks 1

Week 6: Neural Networks 2

Week 7: Reinforcement Learning 1

Week 8: Reinforcement Learning 2

Background Reading List

Tom Mitchell., Machine Learning
Trevor Hastie, Robert Tibshirani, Jerome Friedman., Elements of Statistical Learning
Gareth James, Daniela Witten, Trevor Hastie, Robert Tibshirani., An Introduction to Statistical Learning

If you are planning to purchase books, remember that courses with too few students enrolled will be cancelled. The Department accepts no responsibility for books bought in anticipation of a course.

If you have enrolled on a course starting in the autumn, you can become a borrowing member of the Rewley House library from 1st September and we will try to ensure that as many titles as possible are available in the Library by the start of each term. If you are enrolled on a course starting in other terms, you can become a borrowing member once the previous term has ended.

Certification

Students who register for CATS points will receive a Record of CATS points on successful completion of their course assessment.

To earn credit (CATS points) you will need to register and pay an additional £10 fee per course. You can do this by ticking the relevant box at the bottom of the enrolment form or when enrolling online.

Coursework is an integral part of all weekly classes and everyone enrolled will be expected to do coursework in order to benefit fully from the course. Only those who have registered for credit will be awarded CATS points for completing work at the required standard.

Students who do not register for CATS points during the enrolment process can either register for CATS points prior to the start of their course or retrospectively from between January 1st and July 31st after the current academic year has been completed. If you are enrolled on the Certificate of Higher Education you need to indicate this on the enrolment form but there is no additional registration fee.

Tutor

Dr Cezar Ionescu is Associate Professor of Data Science with the Oxford University Department for Continuing Education. His main interests include functional programming, correctness of scientific computing and machine learning algorithms, and the role of computing science in education.

Course aims

To provide an overview of the field of machine learning by an in-depth presentation of the four or five main algorithms.

Course Objectives

Provide examples, explanations, and demonstrations of logical and bayesian methods, unsupervised learning, neural networks and reinforcement learning.

Teaching methods

Mostly blackboard lectures, accompanied by slides and computer demonstrations.

Learning outcomes

By the end of this course students will be expected to:

1.) understand the main ideas behind machine learning algorithms
2.) understand when a problem can be dealt with using machine learning

3.) determine which kind of machine learning algorithm is likely to be appropriate

Assessment methods

Short Tests

Students must submit a completed Declaration of Authorship form at the end of term when submitting your final piece of work. CATS points cannot be awarded without the aforementioned form.

Level and demands

Most of the Department’s weekly classes have 10 or 20 CATS points assigned to them. 10 CATS points at FHEQ Level 4 usually consist of ten 2-hour sessions. 20 CATS points at FHEQ Level 4 usually consist of twenty 2-hour sessions. It is expected that, for every 2 hours of tuition you are given, you will engage in eight hours of private study.

Credit Accumulation and Transfer Scheme (CATS)

Class