Welcome to the Deep Learning Lectures!
Breaking News
- Exam info added
- Registration in Campusmanagement is still limited - we have requested to change that. Please register there as soon as possible.
- First lecture "official date": Friday April 20, 12:15.
Exam
This year there will be a take home exam instead of a normal exam. The procedure will be the following:
The exam will be online on Monday, the 3rd of August at 1 pm Berlin time and you will have time until Friday, the 7th of August at 1pm (Berlin time) to upload your solutions. However, it should be possible to solve the exam in a couple of hours. That means you will have 96 hours to work on the posed tasks. The tasks will be devised such that everything can be solved in roughly 8 hours.
Exam resit
The exam resit (Nachklausur) is going to be in the same form as the exam. The corresponding dates are: Monday, the 21st of September at 1 pm Berlin time until Friday, the 25th of September at 1pm (Berlin time).
Requirements
As this course is useful for students from many different disciplines, we have no formal requirements. However for a good learning experience, we strongly recommend the following prerequesites:
- Fluent in linear algebra
- Basic statistics
- Python programming
Lectures
The regular teaching hours are Fridays, 12:15-13:45 MEST. However, lectures will be posted on youtube in advance and you can watch them at any time:
https://www.youtube.com/playlist?list=PLqPI2gxxYgMKN5AVcTajQ79BTV4BiFN_0
Individual Lecture Links (optional means will not be relevant for Exam):
- Introduction
- Statistical Estimator Theory
- Neural Networks
- Optimization Methods
- Convolutional Neural Networks
- Autoencoders and PCA
- TICA, time-autoencoders, VAMPnet (optional)
- Recurrent Neural Networks
- Tricks in deep learning
- Probabilistic Models and Restricted Boltzmann Machines
- Variational Autoencoders (optional)
- Generative Adversarial Nets (optional)
Tutorials
Passing the tutorials requirements:
- at least 50% of total achievable points (submit individually, even if you are working in groups)
- at least one "active participation":
- In groups of 3-4 (preferably 4), answer to the corresponding Piazza thread and volunteer
- If you get selected for (online) presentation, send us your solution as PDF (or marked up jupyter notebook if appropriate)
For collaborative work on PDFs and images we recommend overleaf.com and draw.io, respectively.
Whiteboard deadlines:
- Exercises are published every Friday, 13:45 MEST.
- Submission deadline is the subsequent Friday, 12:15 MEST.
The tutors will participate in discussions in the online classroom Piazza and answer questions ongoingly during working hours.
Organization to tutorial groups and first homeworks will be given end of April. More information was sent through Email-notification.
UPDATE: if you did not receive our email with the detailed instructions and an invitation to the Piazza classroom yet, please contact the tutors (Jonas, Leon, Moritz) as soon as possible!