Warsaw University of Technology - Central Authentication System
Strona główna

Pattern Recognition

General data

Course ID: 103C-CSCSN-MSA-EPART
Erasmus code / ISCED: (unknown) / (unknown)
Course title: Pattern Recognition
Name in Polish: Pattern Recognition
Organizational unit: The Faculty of Electronics and Information Technology
Course groups: ( Advanced Courses )--M.Sc.-EITI
( Advanced Courses )-Computer Information System Engineering-M.Sc.-EITI
( Computer Systems and Networks - Advanced )-Computer Systems and Networks-M.Sc.-EITI
( Courses in English )--eng.-EITI
( Technical Courses )---EITI
( Technical Courses )--eng.-EITI
ECTS credit allocation (and other scores): 6.00 Basic information on ECTS credits allocation principles:
  • the annual hourly workload of the student’s work required to achieve the expected learning outcomes for a given stage is 1500-1800h, corresponding to 60 ECTS;
  • the student’s weekly hourly workload is 45 h;
  • 1 ECTS point corresponds to 25-30 hours of student work needed to achieve the assumed learning outcomes;
  • weekly student workload necessary to achieve the assumed learning outcomes allows to obtain 1.5 ECTS;
  • work required to pass the course, which has been assigned 3 ECTS, constitutes 10% of the semester student load.
Language: English
(in Polish) Jednostka decyzyjna:

(in Polish) 103000 - Wydział Elektroniki i Technik Informacyjnych

(in Polish) Kod wydziałowy:

(in Polish) EPART

(in Polish) Numer wersji:

(in Polish) 3

Short description:

The aim of the lecture is to familiarize students with the issue of image recognition. In particular, the following topics will be discussed: general classification of image recognition systems, selected methods and techniques of image recognition as well as issues related to data collection, detection of outliers, dimensionality reduction and assessment of classification quality. The tasks carried out during laboratory exercises and mini-projects will allow the students to become familiar with the selected methods of image classification.

Full description:

The aim of the lecture is to familiarize students with the image recognition, starting from data collection and preprocessing, to selected methods of classification and assessment of the quality of classification.

Lectures:

Introduction: components of the image recognition system; design cycle of creating a classifier; methods of quality assessment of classifiers and classification systems.

Optimal Bayesian Classification: the role of a priori information; the form of the probability density function; optimal Bayes classifier; the risk / loss of the classification decision; decision boundaries of classifiers; compliance of the data distribution with the adopted theoretical distribution.

Nearest Neighbour Methods: Template Matching; minimum distance classifiers; metrics; k-NN classifiers; nearest neighbour search methods; speeding up the search for the nearest neighbour; editing and reduction of the training set.

Linear Classification: linear decision functions; homogeneous space; learning the decision boundary; assembling of individual classifiers' results for more than two classes; support vector method; sequential minimal optimization algorithm.

Dimensionality Reduction: Principal Component Analysis; Fisher's linear classification; multivariate discriminant analysis (MDA).

Clustering: clustering problem; assessment of clustering similarity; k-means class algorithms; bottom-up clustering; graph algorithms.

Neural Networks: basic model of the neuron; single neuron learning algorithms; interpretation of the operation of a single neuron; neural networks; error backpropagation algorithm; convolutional and feedback networks; application examples beyond image recognition.

Markov Models: discrete Markov processes; hidden Markov process; Viterbi's algorithm; Baum-Welsh algorithm for determining the parameters of the Markov system; problems of using Markov models in classification.

Text Searching: the problem of exact and approximate text search; Boyer-Moor algorithm; edit distance; text analysis with the use of nondeterministic and deterministic automata; tree and table of suffixes; Ukkonen's algorithm for constructing a suffix tree; approximate search with suffix trees; neighbourhood generation for search with errors; hash functions for quick searches.

Decision Trees: constructing decision trees - basic CART algorithm; assessment of tree node heterogeneity; stop criteria when building a tree; horizon effect; tree pruning algorithms.

Quality of Classification Improvement: basic problems of designing meta-classifiers; voting schemes; the issue of the independence of classifiers; voting with weights; determination of weights; Bayesian methods of composing the results of classifiers; Behaviour-Knowledge space; methods of constructing sets of weak classifiers (AdaBoost algorithm); use of contextual information in classification; context in OCR systems; use of dictionaries and trigrams.

Laboratories

Nearest Neighbour Classification (warm-up exercise - getting to know the Octave environment).

Bayesian classification with normal distribution and approximation of the probability density distribution with the Parzen window.

Quality of classification improvement - constructing a meta-classifier from the provided classifiers.

Project

Linear classification - reducing dimension with PCA; perceptron algorithm for determining the decision plane; OVO and OVR classifier assemblies; analysis of classification results - confusion matrix.

Neural networks - implementation of the error back propagation algorithm; testing the influence of hyperparameters on the results of classification on the training and validation set.

Bibliography:

  1. Duda R. O., Hart P. E., Stork D. G., Pattern Classification, Wiley-Interscience, 2000
  2. Bishop C. M., Pattern Recognition and Machine Learning, Springer, 2006
  3. Goodfellow I., Bengio Y., Courville A., Deep Learning, MIT Press, 2016 (http://www.deeplearningbook.org/)
  4. Gonzalez R. C., Woods R. E., Digital Image Processing, Pearson, 2018
  5. Press W. H. et al., Numerical Recipes The Art of Scientific Computing, Cambridge University Press, Cambridge 2007 (http://numerical.recipes)

GNU Octave - Scientific Programming Language (https://www.gnu.org/software/octave/)

Learning outcomes:

Knowledge

  • knows basic pattern classification methods
  • knows preliminary data analysis and clustering methods
  • knows basic classifiers ensemble construction methods

Skills

  • can analyze a training set, design a simple classifier and evaluate its quality
  • can, on the basis of a training set assessment, choose a classification method and compute its parameters
  • is able to critically evaluate the solution of the classification problem and propose its improvement

Classes in period "Summer Semester 2023/2024" (in progress)

Time span: 2024-02-19 - 2024-09-30
Selected timetable range:
Navigate to timetable
Type of class:
laboratory, 15 hours, 24 places more information
lectures, 30 hours, 24 places more information
project , 15 hours, 24 places more information
Coordinators: Rajmund Kożuszek
Group instructors: Rajmund Kożuszek
Students list: (inaccessible to you)
Examination: Overall grade
(in Polish) Jednostka realizująca:

(in Polish) 103200 - Instytut Informatyki

Classes in period "Winter Semester 2023/2024" (past)

Time span: 2023-10-01 - 2024-02-18
Selected timetable range:
Navigate to timetable
Type of class:
laboratory, 15 hours, 32 places more information
lectures, 30 hours, 32 places more information
project , 15 hours, 32 places more information
Coordinators: Rajmund Kożuszek
Group instructors: Rajmund Kożuszek
Students list: (inaccessible to you)
Examination: Overall grade
(in Polish) Jednostka realizująca:

(in Polish) 103200 - Instytut Informatyki

Classes in period "Summer Semester 2022/2023" (past)

Time span: 2023-02-20 - 2023-09-30
Selected timetable range:
Navigate to timetable
Type of class:
laboratory, 15 hours, 32 places more information
lectures, 30 hours, 32 places more information
project , 15 hours, 32 places more information
Coordinators: Rajmund Kożuszek
Group instructors: Rajmund Kożuszek
Students list: (inaccessible to you)
Examination: Overall grade
(in Polish) Jednostka realizująca:

(in Polish) 103200 - Instytut Informatyki

Classes in period "Winter Semester 2022/2023" (past)

Time span: 2022-10-01 - 2023-02-19
Selected timetable range:
Navigate to timetable
Type of class:
laboratory, 15 hours, 32 places more information
lectures, 30 hours, 32 places more information
project , 15 hours, 32 places more information
Coordinators: Rajmund Kożuszek
Group instructors: Rajmund Kożuszek, Henryk Rybiński, Muhammad Farhan Safdar
Students list: (inaccessible to you)
Examination: Overall grade
(in Polish) Jednostka realizująca:

(in Polish) 103200 - Instytut Informatyki

Classes in period "Winter Semester 2021/2022" (past)

Time span: 2021-10-01 - 2022-02-22
Selected timetable range:
Navigate to timetable
Type of class:
laboratory, 15 hours, 32 places more information
lectures, 30 hours, 32 places more information
project , 15 hours, 32 places more information
Coordinators: Rajmund Kożuszek
Group instructors: Rajmund Kożuszek
Students list: (inaccessible to you)
Examination: Overall grade
(in Polish) Jednostka realizująca:

(in Polish) 103200 - Instytut Informatyki

Classes in period "Winter Semester 2020/2021" (past)

Time span: 2020-10-01 - 2021-02-19
Selected timetable range:
Navigate to timetable
Type of class:
laboratory, 15 hours, 32 places more information
lectures, 30 hours, 32 places more information
project , 15 hours, 32 places more information
Coordinators: Rajmund Kożuszek
Group instructors: Rajmund Kożuszek
Students list: (inaccessible to you)
Examination: Overall grade
(in Polish) Jednostka realizująca:

(in Polish) 103200 - Instytut Informatyki

Classes in period "Winter Semester 2019/2020" (past)

Time span: 2019-10-01 - 2020-02-21
Selected timetable range:
Navigate to timetable
Type of class:
laboratory, 15 hours, 32 places more information
lectures, 30 hours, 32 places more information
project , 15 hours, 32 places more information
Coordinators: Rajmund Kożuszek
Group instructors: Rajmund Kożuszek
Students list: (inaccessible to you)
Examination: Overall grade
(in Polish) Jednostka realizująca:

(in Polish) 103200 - Instytut Informatyki

Classes in period "Winter Semester 2018/2019" (past)

Time span: 2018-10-01 - 2019-02-17
Selected timetable range:
Navigate to timetable
Type of class:
laboratory, 15 hours, 32 places more information
lectures, 30 hours, 32 places more information
project , 15 hours, 32 places more information
Coordinators: Rajmund Kożuszek
Group instructors: Rajmund Kożuszek
Students list: (inaccessible to you)
Examination: Overall grade
(in Polish) Jednostka realizująca:

(in Polish) 103200 - Instytut Informatyki

Course descriptions are protected by copyright.
Copyright by Warsaw University of Technology.
pl. Politechniki 1, 00-661 Warszawa tel: (22) 234 7211 https://pw.edu.pl contact accessibility statement USOSweb 7.0.0.0-7 (2024-03-18)