Lehr- und Forschungseinheit für Datenbanksysteme
print


Breadcrumb Navigation


Content

Master Seminar "Recent Developments in Deep Learning" (SS 2019)

News:

Organisation

  • Contact: Prof. Dr. Thomas Seidl, Michael Fromm, Evgeniy Faerman, Max Berrendorf
  • Required: Lecture "Knowledge Discovery in Databases I" and Lecture "Machine Learning" or equivalent
    Please indicate in the central registration form in uniworx in which semesters you had attended these courses.
  • Audience: The lecture is directed towards Master students in Mediainformatics, Bioinformatics and Informatics as well as Data Science
  • Registration:: Uniworx

 

Content

The availability of large amounts of data and increased computational power enabled the renaissance of neural networks as an universal paradigm for machine learning.

Neural networks especially profited from the recent developments because the new conditions enabled the training of deep architectures stacking large amounts of hidden layers. In this seminar, students will discuss fundamental works in the quickly evolving area of deep learning and will venture into the state-of-the-art of this exciting development.

In the first part of the seminar participants build teams of 4-5 people and each team is assigned with a general topic. Each team will give an one hour presentation about their topic. Example topics are Recurrent or Convolution Architectures. In the second part of the seminar, each participant presents a particular paper individually.

 

Time and Locations

Event When Location Date
Introduction Fr, 10.00 - 12.00 h Room U 151 (Oettingenstr. 67) 26.04.2019
- 03.05.2019
- 10.05.2019
- 17.05.2019
Short-Presentations I 24.05.2019
Short-Presentations II 31.05.2019
Short-Presentations III 07.06.2019
Presentations I 14.06.2019
Presentations II 21.06.2019
- 28.06.2019
Presentations III 05.07.2019
Presentations IV 12.07.2019
Presentations V 19.07.2019
Presentations VI 26.07.2019

Link to intro slides

Assigned tasks and papers

Additional Information: 

 Presentation:

Templates: