Personen - Universität Osnabrück

Allgemeine Informationen

Veranstaltungsart: Vorlesung
Semester: SoSe 2016
ECTS-Punkte: 4
Veranstaltungsnummer: 8.3269
Details: Link zur Veranstaltung in StudIP

Dozent*innen

Prof. Dr. Gordon Pipa

Tutor*innen

Beschreibung

Prerequisites: Knowledge in neuroinformatics is preferred

The performance of machine learning algorithms is heavily dependent on the representation of the data provided. Data described by means of poor or disadvantageous features often results in a bad performance of the deployed machine learning algorithm. As it is often difficult for humans to hand-code the explanatory factors or features of data beforehand, representation learning aims at learning useful representations of the given data. Various machine learning algorithms can be used in order to perform representation learning. This course introduces the field of representation learning and reviews the major types of feedforward neural networks that can be employed in order to learn representations of data. We will cover topics such as representation learning, artificial neuronal networks, the single and multilayer perceptron, recurrent neuronal networks and deep learning. The seminar will be composed of a mix of lectures that introduce the topics, and presentations given by the students (2 students per presentation) on selected papers that will be uploaded. Every day 10 questions will be discussed to deepen the understanding. Grading will be based on a mixed grade, composed of the oral presentation, active participation during the course, and a written exam coving a selected set of questions from the week. This course will be supplemented by a coding hands-on course in the following week (Title: Hands-on for Representation Learning with Feedforward Neural Networks).

IMPORTANT: The course is limited to 20 participants. Once you have registered, cancellations will only be accepted 4 weeks prior to Aug. 29th. Thereafter, your participation is mandatory.