Course detail

Neural networks and Machine Learning

FSI-VSCAcad. year: 2025/2026

The course introduces basic approaches to Machine Learning and Deep Learning and classical methods used in the field. Practical use of the methods is demonstrated on solving simple engineering problems.

Language of instruction

Czech

Number of ECTS credits

5

Mode of study

Not applicable.

Guarantor

doc. Ing. Pavel Škrabánek, Ph.D.

Department

Institute of Automation and Computer Science (ÚAI)

Entry knowledge

The knowledge of basic relations of the optimization, statistics, graphs theory and programming.

Rules for evaluation and completion of the course

Course-unit credit requirements: submitting a functional software project which uses implementation of selected AI method. Project is specified in the first seminar. Systematic checks and consultations are performed during the semester. Each student has to get through one test and complete all given tasks. Student can obtain 100 marks, 40 marks during seminars (20 for project and 20 for test; he needs at least 20), 60 marks during exam (he needs at least 30).

The attendance at lectures is recommended, at seminars it is obligatory. Education runs according to week schedules. The form of compensation of missed seminars is fully in the competence of a tutor.

Aims

The course objective is to make students familiar with basic resources of Artificial Neural Networks, potential and adequacy of their use in engineering problems solving.
Understanding of basic methods of Artificial Neural Networks and ability of their implementation.

Study aids

Not applicable.

Prerequisites and corequisites

Not applicable.

Basic literature

Goodfellow, I., Bengio, Y., Courville, A.: Deep Learning. MIT Press, 2016. (EN)
Munakata, T.: Fundamentals of the New Artificial Intelligence, Springer-Verlag New York, Inc., 1998. ISBN 0-387-98302-3 (EN)
Sima,J., Neruda,R.: Theoretical questions of neural networks, MATFYZPRESS, 1996, ISBN 80-85863-18-9 (CS)

Recommended reading

B. Kosko: Neural Networks and fuzzy systems. Prentice Hall 1992 (EN)
Bishop, C. M.: Pattern Recognition, Springer Science + Business Media, LLC, 2006, ISBN 0-387-31073-8. (EN)

Classification of course in study plans

  • Programme N-AIŘ-P Master's 1 year of study, summer semester, compulsory

Type of course unit

 

Lecture

26 hod., optionally

Teacher / Lecturer

doc. Ing. Pavel Škrabánek, Ph.D.

Syllabus

1. Introduction to Machine Learning and Soft Computing in the Context of Artificial Intelligence.
2. Evolutionary algorithms I. (genetic algorithms, evolutionary strategies, differential evolution).
3. Evolutionary algorithms II. (grammatical evolution, genetic programming).
4. Selected optimization metaheuristics (HC, HC12, THC, simulated annealing).
5. SWARM Intelligence (PSO, ACO, SOMA).
6. Architectures and classification of neural networks. Perceptron.
7. Feedforward neural networks, single and multilayer networks. ADALINE. Back Propagation Algorithm. Optimization methods used in ANN design.
8. RBF and RCE neural networks. Topologically organized neural networks (competitive learning, Kohonen maps).
9. Cluster analysis. Task dimension reduction. Principal component analysis. LVQ neural networks, neural networks ART.
10. Associative neural networks (Hopfield, BAM), behavior, state diagram, attractors, learning. and Neocognitron.
11. Deep Neural Network. CNN. Transfer Learning.
12. Spiking neural Network.
13. Case studies. Deterministic chaos and its control.

Computer-assisted exercise

26 hod., compulsory

Teacher / Lecturer

doc. Ing. Pavel Škrabánek, Ph.D.

Syllabus

Seminars related to the lectures in the previous week. Solution Topics:
- Implementation of basic metaheuristics
- solving global optimization problems
- use of global optimization toolbox
- use of deep neural network toolbox
- creation of nonlinear models using neural networks
- deep learning in computer vision for image classification
- detection of objects in Image using Deep Learning (R-CNN)
- Semantic Image Segmentation using Deep Learning (SegNet)
- validation of CNN learning and control of learned networks using deep dream method