Lab Course (Master level)

Deep Learning Course, Control Section
Dr. J. Boedecker, J. T. Springenberg

  • Announcements:
    • first meeting: 21.10.2016
  • Dates
    • Friday, 14:00 - 16:00, building 082 - Kinohoersaal / computer pools (as announced)
  • Language
    • English

Overview

Deep learning has brought a revolution to AI research. A good understanding of the principles of deep networks and experience in training them has become one of the main assets for successful research and development of new technology in machine learning, computer vision, and robotics.

In this course, which will be jointly organized by the computer vision group, the robotics group, and the two machine learning groups, we want to teach students the practical knowledge that is needed to do research with deep learning in any of these fields. The course starts with some introductory lectures, continues with first some smaller and then larger projects. You must work in teams of 2-3 persons. There will be a final presentation of your project results at the end of the semester.

Presentations

  • 1 Lecture: Introduction (pdf)
  • 2 Lecture: Implementing CNNs, the presentation I gave was based on Andrej's great slides on implementing CNNs from here and the starting slides I used can be found here
  • Assignment 2 slides for Track 2 can be found here
  • Assignment 3 slides for Track 2 can be found here
  • NEW Assignment 4 slides for Track 2 can be found here the exercise sheet can be downloaded here

Assignments

* You can find the course assignments on github feel free to fork them (and of course pull requests for improvements will be considered ;))

Further Material

  • Draft of a recent book on deep learning by Yoshua Bengio, Ian Goodfellow and Aaron Courville (link)
  • Course on Neural Networks by Hugo Larochelle (link)
  • Deep learning tutorial by Kyunghyun Cho (pdf)
  • Stanford course on deep learning and unsupervised feature learning (link)
  • A great recent course by Andrej Karpathy on convolutional neural networks (link)

If you need cudnn please get it from NVIDIA directly. On the pool machines you can also use the cudnn version stored in the mllect home by exporting your library path like this:

  • export LD_LIBRARY_PATH=/home/mllect/new_cudnn/lib:$LD_LIBRARY_PATH

or, if you need the old cudnn version

  • export LD_LIBRARY_PATH=/home/mllect/old_cudnn/lib:$LD_LIBRARY_PATH