courses:be5b33rpz:lectures:start [CourseWare Wiki]

Summary

This course introduces statistical decision theory and surveys canonical and advanced classifiers such as perceptrons, AdaBoost, support vector machines, and neural nets.

Basic info

Winter semester 2022/2023

Where and when: KN:E-301 at Building G, Karlovo namesti, Monday 16:15-17:45

Lectures will be streamed on YouTube, follow this link + select KN:E-301@CVUTFEL-live. Recorded lectures will be available on a YouTube playlist before Tuesday morning. For online feedback, connect via zoom link: https://feectu.zoom.us/j/6135640703

Teaching: Jiří Matas (JM) matas@cmp.felk.cvut.cz, Ondřej Drbohlav (OD) drbohlav@cmp.felk.cvut.cz

Lecture plan 2022/2023

Week	Date	Lect.	Slides	Topic	Wiki	Additional material
1	19.9.	JM	pdf	Introduction. Basic notions. The Bayesian recognition problem	Machine_learning Naive_Bayes_classifier	some simple problems
2	26.9.	OD	pdf	Non-Bayesian tasks	Minimax
3	3.10.	OD	pdf	Parameter estimation of probabilistic models. Maximum likelihood method	Maximum_likelihood
4	10.10.	OD	pdf	Nearest neighbour method. Non-parametric density estimation.	K-nearest_neighbor_algorithm
5	17.10.	OD	pdf	Logistic regression	Logistic_regression
6	24.10.	OD	pdf	Classifier training. Linear classifier. Perceptron.	Linear_classifier Perceptron
7	31.10.	JM	pdf	SVM classifier	Support_vector_machine	demo
8	7.11.	JM	pdf	Adaboost learning	Adaboost
9	14.11.	JM	pdf	Neural networks. Backpropagation	Artificial_neural_network
10	21.11.	JM	pdf	Cluster analysis, k-means method	K-means_clustering K-means++
11	28.11.	JM	pdf	EM (Expectation Maximization) algorithm.	Expectation_maximization_algorithm	Hoffmann,Bishop, Flach
12	5.12.	JM	pdf	Feature selection and extraction. PCA, LDA.	Principal_component_analysis Linear_discriminant_analysis	Optimalizace (CZ): PCA slides, script 7.2
13	12.12.	JM	pdf	Decision trees.	Decision_tree Decision_tree_learning	Rudin@MIT
14	9.1.	JM	Basic notions recapitulation, links between methods, answers to exam questions	The lecture will not take place at the standard time, I'll be answering question by email or zoom link

Recommended literature

Duda R.O., Hart, P.E.,Stork, D.G.: Pattern Classification, John Willey and Sons, 2nd edition, New York, 2001
Schlesinger M.I., Hlaváč V.: Ten Lectures on Statistical and Structural Pattern Recognition, Springer, 2002
Bishop, C.: Pattern Recognition and Machine Learning, Springer, 2011
Goodfellow, I., Bengio, Y. and Courville, A.: Deep Learning, MIT Press, 2016. www

Assessment (zápočet)

Conditions for assessment are in the lab section.

Exam

Only students who receive all credits from the lab work and are granted the assessment (“zápočet”) can be examined.
The labs contribute 50% to your final evaluation, the written part of the exam contributes 40% and the oral part 10%.
A threshold for passing the exam is set, usually between 5-10 points (out of 40), depending on the complexity of the test.
Your grade chances after the written exam test and before the oral exam are illustrated in the image below. Beware: The scheme is approximate only, but illustrates how the oral exam influences the final grade.
The questions used in the test are available here (if one can solve these questions, one will likely do well in the exam)
The oral starts approximately 2 hours after the end of the test, as soon as the tests are graded. Alternatively, the oral might take place the following day, if the number of tests to correct is high.
The oral part is compulsory! Note that if during the oral examination it becomes clear that the student cannot explain how they obtained a result in the written exam, the examiner may re-assess the grading of the written exam.
Example oral exam questions are available here.

Table of Contents