Learning from Data (Level 10) (VS1) (U02405)

? Credit Points : 10 ? SCQF Level : 10 ? Acronym : INF-4-LFD-V

Since the early days of AI, researchers have been interested in making computers learn, rather than simply programming them to do tasks. This is the field of machine learning. The main area that will be discussed is supervised learning, which is concerned with learning to predict an output, given inputs. For example, we might wish to classify a handwritten digit as one of the numbers 0 through 9, given an input image of the digit. We will compare and contrast different algorithms for supervised learning tasks. One of the key issues is that of generalization, i.e. how good our performance will be on new input patterns. A second area of study is unsupervised learning, where we wish to discover the structure in a set of patterns; there is no output "teacher signal".

The field of learning from data is of increasing importance, given the large volume of data that can now be collected in many different domains, ranging from time recordings of multiple neurons from regions of the brain, to point-of-sale information from supermarkets; the task is to make sense of this data by understanding its structure and using it to make predictions.

The primary aim of the course is to provide you with a set of practical tools that you can apply to solve real-world problems in machine learning, coupled with an appropriate, principled approach to formulating a solution.

Entry Requirements

? This course is only available to part year visiting students.

? This course is a variant of the following course : U01949

? Pre-requisites : Successful completion of Year 3 of an Informatics Single or Combined Honours Degree, or equivalent by permission of the School. Familiarity with elementary mathematics, including algebra and calculus is essential. A reasonable knowledge of computational, logical, geometric and set-theoretic concepts is assumed. Knowledge of vectors and matrices, together with a basic grasp of probability and partial differentiation, is strongly recommended. This course is only available to part-year visiting students who are only in Edinburgh for Semester 1.

? Prohibited combinations : Learning from Data (Level 11)

Subject Areas

Home subject area

Bioinformatics, (School of Informatics, Schedule O)

Other subject areas

Computational Systems Biology, (School of Informatics, Schedule O)

Learning from Data, (School of Informatics, Schedule O)

Intelligent Robotics, (School of Informatics, Schedule O)

Database Systems, (School of Informatics, Schedule O)

Speech Processing, (School of Informatics, Schedule O)

Life Science Computing, (School of Informatics, Schedule O)

Language Processing, (School of Informatics, Schedule O)

Delivery Information

? Normal year taken : 4th year

? Delivery Period : Semester 1 (Blocks 1-2)

? Contact Teaching Time : 3 hour(s) per week for 10 weeks

First Class Information

Date	Start	End	Room	Area	Additional Information
20/09/2007	17:10	18:00	Lecture Theatre 3, Appleton Tower	Central

All of the following classes

Type	Day	Start	End	Area
Lecture	Monday	17:10	18:00	Central
Lecture	Thursday	17:10	18:00	Central

? Additional Class Information : M-F 0900-1700 as arranged.

Summary of Intended Learning Outcomes

This course involves some implementation work using MATLAB, and contains also significant theoretical work involving areas of mathematics other than logic. The course details the specific applications of various basic mathematical techniques to areas of pattern recognition and processing, the aim being that at the end of the course, participants should have a good understanding and ability to use these techniques in practice. The course aims to foster a systematic approach to experiments.

A fairly deep understanding of machine learning, its aims and limitations, from a viewpoint of modelling. Students will learn how to approach a problem and assess from a variety of methods an appropriate solution.

A set of practical methods, each coupled with an algorithm in MATLAB, will enable the student to be able to program solutions to real world machine learning problems, using the toolbox of practial methods presented in the lectures.

A basic understanding of how mathematics can be applied to solving real world problems. A good understanding of assessing scientific hypotheses. MATLAB programming.