PA196 Pattern Recognition

Faculty of Informatics
Autumn 2018
Extent and Intensity
2/2/0. 4 credit(s) (plus extra credits for completion). Type of Completion: zk (examination).
Teacher(s)
doc. Ing. Vlad Popovici, PhD (lecturer), doc. RNDr. Petr Matula, Ph.D. (deputy)
Mgr. Anna Pačínková, Ph.D. (assistant)
Guaranteed by
doc. RNDr. Petr Matula, Ph.D.
Department of Visual Computing – Faculty of Informatics
Supplier department: Department of Visual Computing – Faculty of Informatics
Timetable
Thu 14:00–15:50 C416
  • Timetable of Seminar Groups:
PA196/01: Thu 16:00–17:50 A219, V. Popovici
Prerequisites
At least working knowledge of statistics/probabilities, linear algebra and mathematical analysis are required
Course Enrolment Limitations
The course is offered to students of any study field.
Course objectives
By successful completion of the course, the students will: (i) have solid understanding of the principles of pattern recognition; (ii) master the main methods for model performance estimation; (iii) have a good grasp of different parametric and non-parametric methods for classification; (iv) understand the main classification methods; (v) have a working knowledge of kernel methods; (vi) understand and master the performance estimation techniques (vii) have hands-on experience of using pattern recognition methods in computer vision and biomedical applications.
Syllabus
  • 1. Introduction: problem setting; distances, metrics, similarity; Bayesian decision theory 2. Non-parametric methods: density estimation; nearest neighbor methods 3. Classification performance: performance criteria; performance estimation; confidence intervals; classifier comparison 4. Linear discriminants: decision surfaces; parameter optimization; shrinkage, penalized regression, optimal separating hyperplanes; SVM 5. Ensemble methods: fusion of labels or continuous outputs; parallel and cascaded systems; bagging and boosting; AdaBoost 6. One-class classifiers: Gaussian mixtures; support vector density estimators; outlier detection 7. Dimensionality reduction: feature selection; PCA, ICA, non linear PCA 8. Unsupervised learning/clustering: principles; mixture of densities; hierarchical clustering
Literature
    required literature
  • HASTIE, Trevor, Robert TIBSHIRANI and J. H. FRIEDMAN. The elements of statistical learning : data mining, inference, and prediction. 2nd ed. New York, N.Y.: Springer, 2009, xxii, 745. ISBN 9780387848570. info
    recommended literature
  • KUNCHEVA, Ludmila. Combining pattern classifiers: methods and algorithms. John Wiley & Sons, 2004. ISBN 0-471-21078-1. info
  • DUDA, Richard O., David G. STORK and Peter E. HART. Pattern classification. 2nd ed. New York, N.Y: John Wiley & Sons, 2001, xx, 654. ISBN 0471056693. info
    not specified
  • HEIJDEN, Ferdinand van der, Robert DUIN, Dick de RIDDER and David M J TAX. Classification, Parameter Estimation and State Estimation: An Engineering Approach Using MATLAB. 2004. ISBN 978-0-470-09013-8. info
Teaching methods
Lectures; group work on specific topics; laboratory classes.
Assessment methods
Written and practical exams; bonuses for attendance and involvement
Language of instruction
English
Further Comments
Study Materials
The course is taught annually.
The course is also listed under the following terms Autumn 2014, Autumn 2015, Autumn 2016, Autumn 2017, Autumn 2019.
  • Enrolment Statistics (Autumn 2018, recent)
  • Permalink: https://is.muni.cz/course/fi/autumn2018/PA196