Machine Learning: Unsupervised Methods (with Problem Based Learning)
This course is given in a hybrid of inverted classroom and problem based learning. The course starts with a two-week introduction into unsupervised methods of machine learning, providing an overview. The students then work in groups of about 4 on realistic problems that can be solved with these methods. In the first week of a problem, they develop hypotheses and strategies for a solution and identify which methods they want to learn. Then the course agrees on a method to focus on theoretically, which will then be done in an inverted classroom format. The students then try to solve the problem and present their results in a short video talk with slides. Thus the students will not only learn about machine learning but also soft skills.
This course covers a variety of unsupervised methods from machine learning such as principal component analysis, independent component analysis, vector quantization, clustering, Bayesian theory and graphical models. These are more classical methods, this course does not cover deep learning methods by default, although it is OK to use such methods to solve the problems, but they will not be discussed in the theoretical sessions.
- know a number of important unsupervised learning methods,
- can discuss and decide which of the methods are appropriate for a given data set,
- understand the mathematics of these methods,
- know how to implement and apply these methods in python,
- have gained experience in organizing and working in a team,
- know problem solving strategies like brain storming,
- can communicate about all this in English.
Exam (Prüfungsformen):
50% of the grade come from the average group performance on solving the problems. 10% come from the video presentations every student has to prepare, taking into account slides and presentation style, this is an individual grade of the presenter. 40% come from a digital quiz (or oral exam, if the number of participants is low) about the theory of the methods covered. Thus 50% of the grade are individual, 50% come from the group. Individual as well as group score, both must be passed in order to pass the course. If you fail in one of them you fail in the whole course. In addition you can gain up to 4 bonus points for being voted for as a 'most valuable player (MVP)' on a project. Since the exam is distributed over the semester, students (at least of (Applied) Computer Science) must register for it at the beginning of the semester.
Condition for granting the credit points (Voraussetzungen für die Vergabe von Kreditpunkten): Continuous participation, successful group work (scoring at least 50/100 points), successful individual contribution (scoring at least 50/100 points in presentation and exam).
Max. number of participants: 30
Teaching Material:
Lecturers
Prof. Dr. Laurenz WiskottLecturer |
(+49) 234-32-27997 laurenz.wiskott@ini.rub.de NB 3/29 |
Details
- Course type
- Project seminar
- Credits
- 9 CP
- Term
- Winter Term 2024/2025
- E-Learning
- moodle course available
Dates
- Project seminar
-
Takes place
every week on Tuesday from 10:30 to 13:45 in room ID 04/471 + 459.
First appointment is on 08.10.2024
Last appointment is on 28.01.2025
Requirements
The mathematical level of the course is mixed but generally high, including calculus (functions, derivatives, integrals, differential equations, ...), linear algebra (vectors, matrices, inner product, orthogonal vectors, basis systems, ...), and a bit of probability theory (probabilities, probability densities, Bayes' theorem, ...). Programming is done in Python, thus the students should have a basic knowledge of that as well, or at least be fluent in another programming language.
The Institut für Neuroinformatik (INI) is a central research unit of the Ruhr-Universität Bochum. We aim to understand the fundamental principles through which organisms generate behavior and cognition while linked to their environments through sensory systems and while acting in those environments through effector systems. Inspired by our insights into such natural cognitive systems, we seek new solutions to problems of information processing in artificial cognitive systems. We draw from a variety of disciplines that include experimental approaches from psychology and neurophysiology as well as theoretical approaches from physics, mathematics, electrical engineering and applied computer science, in particular machine learning, artificial intelligence, and computer vision.
Universitätsstr. 150, Building NB, Room 3/32
D-44801 Bochum, Germany
Tel: (+49) 234 32-28967
Fax: (+49) 234 32-14210