The lecture introduces the basic concepts of deep learning, in particular different popular architectures including restricted Boltzmann machines, convolutional neural networks, stacked autoencoders (as concerns feedforward models), and long short term memory models and recursive networks (as concerns recurrent models). Different tricks for training are introduced including optimiziation schemes such as ADAM, and training modi such as drop-out, and training set enhancement. Popular applications such as for image recognition will be discussed. On a practical side,mostly PyTorchwill be used.
NOTE: this lecture will most likely take place as a compact couse at the end of September, provided the situation allows it at that time.
Programmierkenntnisse (Python oder vergleichbar), Grundlagen Mathematik, Grundlagen maschinelles Lernen (etwa Neuronale Netze oder Einführung maschinelles Lernen)
Frequency | Weekday | Time | Format / Place | Period |
---|
Module | Course | Requirements | |
---|---|---|---|
39-M-Inf-DL Deep Learning | Deep Learning | Graded examination
|
Student information |
The binding module descriptions contain further information, including specifications on the "types of assignments" students need to complete. In cases where a module description mentions more than one kind of assignment, the respective member of the teaching staff will decide which task(s) they assign the students.
Students are able distinguish different key architectures for deep learning, and they know the learning scenarios where these architectures can be used. The students know how to train a deep architecture, both, as regards the algorithmic pipeline and possible forms of implementation using modern systems.
A corresponding course offer for this course already exists in the e-learning system. Teaching staff can store materials relating to teaching courses there: