1. Machine Learning (Extended) Dr. Ata Kaban
Algorithms to enable computers to learn
Learning = ability to improve performance automatically through experience
Experience = previously seen examples
Interdisciplinary field AI
Probability & Statistics
Information theory
Philosophy
Control theory
Psychology
Neurobiology, etc

2. What is the Learning Problem?

3. Example: Which word a person is thinking about?
FMRI brain activity data:
Source: Tom Mitchell's research pages

4. Example: Find a specified object

7. s 1
s 2
s 3
s 4
x 1
x 2
x 3
x 4
a 11
a 12
a 13
a 14

8. What is the Learning problem?
Learning = improving with experience at some task
Improve at task T
With respect to performance measure P
Based on experience E
Example: Learning to play checkers
T: play checkers
P: % of games won in world tournament
E: opportunity to play against self

9. Example: Learning to recognise faces
T: recognise faces
P: % of correct recognitions
E: opportunity to make guesses and being told what the truth was
Example: Learning to find clusters in data
T: finding clusters
P: compactness of the groups detected
E: opportunity to see a large set of data

10. Types of training experience
Direct or indirect
With a teacher or without a teacher
An eternal problem: is the training experience representative of the performance goal? – it needs to be.

11. Forms of Machine Learning
Supervised learning: uses a series of examples with direct feedback
Unsupervised learning: no feedback
Reinforcement learning: indirect feedback, after many examples
Q: For the examples given, can you distinguish which type of learning they belong to?

12. Some achievements of ML
Programs that can:
Recognize spoken words
Predict recovery rates of pneumonia patients
Detect fraudulent use of credit cards
Drive autonomous vehicles
Play games like backgammon – approaching the human champion!

13. Focus of the module Understanding the fundamental principles
Types of ML tasks
General algorithms and how they work
Which method is good for what and why
What ML methods can and cannot do
Open research problems
This module is NOT a course on teaching to use a particular software package

14. Syllabus
1.Overview of machine learning. Basic notions, literature
2.Supervised learning Generative methods Discriminative methods Computational learning theory basics Boosting and ensemble methods 3.Unsupervised learning Clustering Learning for structure discovery 4.Reinforcement learning basics 5.Topics in learning from high dimensional data and large scale learning

15. Literature Machine Learning (Mitchell)
A first course in Machine Learning (Rogers & Girolami)
Support Vector Machines and Other Kernel-Based Learning Methods (Cristianini, Shawe-Taylor)
Modelling the Web (Baldi, Smyth)
Artificial Intelligence … (Russell, Norvig)
+math refreshers on the ML module's website

16. Machine Learning & Machine Learning Extended -- Administratrivia --
Ata Kaban
2014/15

17. Assessment
Machine Learning Extended: 40% Coursework; 60% exam.
Coursework = Assignments of Type 1 and Type 2:
Type 1: Paper based problems. These will be assessed in one Class Test (15%) and one take-home assignment (5%). For each, the date and deadline date will be communicated one week in advance.
Type 1: Computer based problems. These consist of 4 pieces of work, handed out throughout the term, with deadline at the end of term.
Machine Learning: 20% Coursework; 80% exam.
Coursework:
Type 1 in the form of a class test (15%)
For the remaining 5% you can choose between a Type 1 assignment or any of the four Type 2 assignments above.
Non-assessed formative exercise sheets will be given out frequently. These are highly recommended for practice.

18. Classes, web site 2 hours / week
Some are lectures and some are exercise classes.
Module home page:
The content currently there is from last year, and the page will be updated as we go along. However it gives you a good idea of what to expect in terms of content, level of difficulty, types of assignments etc.
Contains some math refreshers you might find useful: Linear Algebra & Probability Theory for Machine Learning

19. Office hours
The time for my weekly office hours is communicated on my timetable (watch for possible changes):
Location: UG32
What office hours are and aren’t for
Yes: ask me concrete questions to clarify something that has not been clear to you from the lecture
Yes: seek advice on your solutions to the given exercises
Yes: seek advice on further readings on related material not covered in the lecture
No: ask me to solve the exercises
No: ask me to repeat a lecture