Course details

Machine Learning and Recognition

SUR Acad. year 2021/2022 Summer semester 5 credits

Current academic year

The tasks of classification and pattern recognition, basic schema of a classifier, data and evaluation of individual methods, statistical pattern recognition, feature extraction, multivariate Gaussian distribution,, maximum likelihood estimation, Gaussian Mixture Model (GMM), Expectation Maximization (EM) algorithm, linear classifiers, perceptron, Gaussian Linear Classifier, logistic regression, support vector machines (SVM), feed-forward neural networks, convolutional and recurrent neural networks, sequence classification, Hidden Markov Models (HMM). Applications of the methods of speech and image processing.


Course coordinator

Language of instruction



Examination (written+oral)

Time span

  • 26 hrs lectures
  • 13 hrs exercises
  • 13 hrs projects

Assessment points

  • 60 pts final exam (written part)
  • 15 pts mid-term test (written part)
  • 25 pts projects




Subject specific learning outcomes and competences

The students will get acquainted with the problem of machine learning applied to pattern classification and recognition. They will learn how to apply basic methods in the fields of speech processing and computer graphics. They will understand the common aspects and differences of the particular methods and will be able to take advantage of the existing classifiers in real-situations.
The students will get acquainted with python libraries focused on math, linear algebra and machine learning. They will also improve their math skills (probability theory, statistics, linear algebra) a programming skills. The students will learn to work in a team.

Learning objectives

To understand the foundations of machine learning with the focus on pattern classification and recognition. To learn how to apply basic algorithms and methods from this field to problems in speech and image recognition. To conceive basic principles of different generative an discriminative models for statistical pattern recognition. To get acquainted with the evaluation procedures.

Why is the course taught

Recent years witnessed a boom of machine learning or pattern recognition applications. More and more devices can be controlled using voice or gestures. Digital cameras automatically detect faces in the captured images in order to automatically focus or somehow react on it. Virtual agents in mobile devices can recognize speech and search for relevant answers to our queries. The quality of the current systems for automatic recognition of a person's identity from voice recording or from face photo already significantly exceed the human abilities.

In this class, the students should learn how these technologies work. They will learn about the basic algorithms and models, which, using some training examples, automatically learn to recognize nontrivial patterns in audio recordings, images or other signals or input data.

Recommended prerequisites

Prerequisite knowledge and skills

Basic knowledge of the standard math notation.

Study literature

  • Bishop, C. M.: Pattern Recognition, Springer Science + Business Media, LLC, 2006, ISBN 0-387-31073-8.
  • Hart, P. E., Stork, D. G.:Pattern Classification (2nd ed), John Wiley & Sons, 2000, ISBN: 978-0-471-05669-0.
  • Ian Goodfellow, Yoshua Bengio, and Aaron Courville. 2016. Deep Learning. The MIT Press.

Fundamental literature

Syllabus of lectures

  1. The tasks of classification and pattern recognition, the basic schema of a classifier, data sets and evaluation
  2. Probabilistic distributions, statistical pattern recognition
  3. Generative and discriminative models
  4. Multivariate Gaussian distribution, Maximum Likelihood estimation,
  5. Gaussian Mixture Model (GMM), Expectation Maximization (EM)
  6. Feature extraction, Mel-frequency cepstral coefficients.
  7. Application of the statistical models in speech and image processing.
  8. Linear classifiers, perceptron
  9. Gaussian Linear Classifier, Logistic regression
  10. Support Vector Machines (SVM), kernel functions
  11. Neural networks - feed-forward, convolutional and recurrent
  12. Hidden Markov Models (HMM) and their application to speech recognition
  13. Project presentation

Syllabus of numerical exercises

Lectures will be immediately followed by demonstration exercises (1h weekly) where examples on data and real code will be presented. Code and data of all demonstrations will be made available to the students.

Syllabus - others, projects and individual work of students

  • Individually assigned projects

Progress assessment

  • Mid-term test - up to 15 points
  • Project - up to 25 points
  • Written final exam - up to 60 points

Controlled instruction

The evaluation includes a mid-term test, individual project, and the final exam. The mid-term test does not have a correction option, the final exam has two possible correction terms

Course inclusion in study plans

Back to top