FEKT-MSTUAcad. year: 2017/2018
The field of machine learning is concerned with the question of how to construct computer programs that automatically improve with experience. The goal of the subject is to present the key algorithms and theory that form the core of machine learning. Machine learning is mathematical-logical base in many fields including artificial intelligence, pattern recognition or data mining. The main attention is given on classification and optimization tasks.
Learning outcomes of the course unit
The graduate is able to
- design own solution of a classification task
- pre-process data, including feature selection
- estimate quality of selected model
- justify rightness of suggested solution
- design own solution of optimization task
- select appropriate search heuristic for given problem
The subject knowledge on the Bachelor´s degree level is requested namely in mathematics, statistics and probability theory.
Recommended optional programme components
Recommended or required reading
Mitchell, Tom M. Machine learning. Boston : McGraw-Hill, 1997. 414 s. McGraw-Hill series in computer science. ISBN 0-07-042807-7. (EN)
Honzík P.: Strojové učení. Elektonická skripta VUT. (CS)
Planned learning activities and teaching methods
Techning methods include lectures and computer laboratories. Students have to write a single project during the course.
Assesment methods and criteria linked to learning outcomes
A group project (40 pts) and a final exam (60 pts) are evaluated during the Machine Learning course. For successful pass the course, obtaining of at least half of available points is required in both mentioned parts.
Language of instruction
1. Machine learning paradigms. Terminology. Concept learning. Basics of information theory.
2. Statistics in machine learning.
3. Instance based learning.
4. Decision trees.
5. Model performance estimation.
6. Loss functions. Pre-processing 1.
7. Pre-processing 2.
8. Genetic algorithms. Differential evolution. Ant colony optimization.
9. Bayesian learning.
10. Linear regression. Discriminant analysis. Support vector machines.
11. Meta learning, ensemble methods.
12. Unsupervised learning.
The aim are large knowledge in machine learning with emphasise on classification and optimisation tasks.
Specification of controlled education, way of implementation and compensation for absences
Only registration and submitting of the project are obligatory to qualify for examination.