**P33ROD - Pattern
Recognition for doctoral students**

Summer semestr 2006/2007, 2 hours lectures, personal assignment equivalent to 2 hours per week.

The lecture is given on Thursdays from 9:15 to 10:45 in the seminar room G205, building G of the Faculty of Electrical Engineering at the Czech Technical University Campus at Karlovo náměstí. The lectures are given in English because there are students enrolled who do not understand Czech.

The subject P33ROD je intended as an introductory one for doctoral students who did not pass the subject 33RPZ Pattern Recognition which is taught by Doc. Ing. Jiří Matas, PhD. in the magister study program. For doctoral students knowing basic pattern recognition, a more advanced subject P33ROZ is available which is taught in the same semester.

Lecturers:

** prof. Ing. Václav Hlaváč,
CSc., prof. Ing. Vladimír Mařík, DrSc.**

Katedra kybernetiky, Fakulta elektrotechnická ČVUT

hlavac@cmp.felk.cvut.cz, marik@fel.cvut.cz

Exercises instructor: **prof.
Ing. Václav Hlaváč, CSc.**

Students have two options how to pass exercises.

- Participate in the exercises taught for master students in the subject 33RPZ. Make your arrangements with the instructors of the subject. The content of exercises can be modified according to previous experience of the student and according to student's research interest.
- Solve an individual assignment which is related to the subject. This option has to be negotiated in advanced with V. Hlaváč.

Praha, February 26, 2007

**Plan of the lectures**

Week of a semester |
Lecturer |
Date |
Topic of the lecture |

1 | V. Hlaváč | 1.03.2007 | Introduction. Basic concepts. Formulation of the basic tasks solved in pattern recognition. Rehearsal of basic knowledge from probability theory and statistics. |

2 | V. Hlaváč | 8.03.2007 | Bayesian task. Overview of non-Bayesian task. |

3 | V. Mařík | 15.03.2007 | Structural pattern recognition. (Lecture will be given in Czech, non-Czech speaking students will be given replacement training by prof. V. Mařík) |

4 | V. Hlaváč | 22.03.2007 | Two special statistical models. Conditional independence of features. Gaussian models. Strightening of the feature space. |

29.03.2007 |
No lecture because V. Hlaváč is on business
trip to France.. |
||

5 | V. Hlaváč | 5.04.2007 | Estimation of probabilistic models. Parametric and nonparametric methods. |

6 | V. Hlaváč | 5.04.2007 | Learning in pattern recognition. VC dimension. Estimate of the needed length of the training sequence. (extra lecure from 12:45 to 14:15 to compensate for the canceled lecture on 29.03.2007) |

7 | V. Hlaváč | 12.04.2007 | Linear classifiers and its learning. |

8 | V. Hlaváč | 19.04.2007 | Support vector machines classifiers (SVM). Kernel methods. |

9 | V. Hlaváč | 26.04.2007 | Participation at the One-day colloquium of the CMP (9:00 - 16:00) |

10 | V. Hlaváč | 3.05.2007 | Unsupervised learning. Cluster analysis. EM (Expectation Maximization) algorithm. |

11 | V. Hlaváč | 10.05.2007 | Recognition of Markovian sequences. |

12 | V. Hlaváč |
17.05.2007 |
Graphical models. |

24.05.2007 |
No lecture because V. Hlaváč will be on a
business trip to Japan. |
||

13 | V. Hlaváč | 31.05.2007 | Feature ranking and selection. Experimental evaluation of the classifiers. Receiver operator curve (ROC). |

14 | V. Hlaváč | 7.06.2007 | Relation between statistical and structural pattern recognition. |

Recommended literature:

- Schlesinger M.I., Hlaváč V.:
*Ten lectures on statistical and structural pattern recognition*, Kluwer Academic Publishers, Dordrecht, The Netherlands, 2002. Edition in Czech Vydavatelství ČVUT Praha, 1999. - Duda R.O., Hart, P.E.,Stork, D.G.:
*Pattern Classification*, John Willey and sons, 2nd edition, New York, 2001. - Bishop, C.M:
*Pattern Recognition and Machine Learning,*Springer, New York, 2006.

Students of the subject:

- Lukáš Cerman
- Petr Ježdík, abroad in Portugal for a semester, does not attend lectures.
- Hynek Kružík, abroad for a semester, does not attend lectures.
- Lubomír Říha
- Tanja Schilling
- Zdenka Uhríková

Presentations (It is likely that presentation will be updated before the lecture):

- Introduction to pattern recognition
- Bayesian approach to pattern recognition
- Non-Bayesian tasks
- Linear programning (Lecture A, Lecture B)
- Two statistical models
- Learning in pattern recognition
- Linear classifier
- Support Vector Machines (SVMs)
- Markovian sequences (Rabiner's tutorial)
- Relation between statistical and structural pattern recognition (in Czech)

*Poslední změna 26.02. 2007
V. Hlaváč,
*hlavac@fel.cvut.cz