Deep Learning: Basic Course Information
Scope: Introduction to deep learning methods. We will also cover some basics of shallow Machine Learning methods whenever they are helpful to understand deep learning aspects.
Format: The course has 2 lecture hours per week and homework with 2 tutorial hours per week (2+2 semester week hours) and counts as 5 credit points. Lectures will be recorded and viewable on demand online.
Course registration:
- All students should be registered in WhiteBoard (here) in order to receive relevant information.
- All FU Berlin students should be registered in CampusManagement in order to be able to receive a grade. Students outside Math/CS will likely not see this course in their CampusManagement - in this case please ask your Studienbüro to register you.
Tutorials: We offer three weekly tutorials over CampusWire live room:
- Mondays 12:15 - 13:45
- Tuesdays 12:15 - 13:45
- Wednesdays 10:15 - 11:45
Please sign up for one of these time slots here.
Requirements: To follow this course the following skills are required and helpful:
- Essential: good practice with linear Algebra (we will frequently do calculations with Matrix-Vector operations and Matrix decompositions)
- Essential: good practice with Python programming.
- Helpful: Basic knowledge in Statistics, Optimization
- Helpful: Experience with Python machine learning frameworks such as sklearn, PyTorch, TensorFlow or JAX.
Lectures
Lectures will be recorded and you can watch them at any time. Most lectures will be equal or similar to the 2020 lectures -- you can find the full 2020 playlist here. Some of these lectures will be changed, updated and there will be some new lectures.
This is the official (exam-relevant) list of lectures for the course which will be extended throughout the semester:
- Lecture 1 (Apr 16): Introduction
- Lecture 2 (Apr 23): Regression and Statistical Estimator Theory
- Lecture 3 (Apr 30): Neural Networks Introduction
- Lecture 4 (May 7): Stochastic Gradient Descent Methods
- Lecture 5 (May 14): Convolutional Neural Nets
- Lecture 6 (May 21): Manifold Learning 1: PCA and Autoencoders
- Lecture 7 (May 28): Manifold Learning 2: TICA, time-Autoencoders, VAMPnets
- Lecture 8 (Jun 4): Recurrent Neural Networks
- Lecture 9 (Jun 11): Tools and Tricks
- Lecture 10 (Jun 18): Probabilistic Models
- Lecture 11 (Jun 25): Generative Models I
- Lecture 12 (Jul 2): Generative Models II
- Lecture 13 (Jul 9): Bonus Lecture on Reinforcement Learning, Normalizing Flows and Physics-constrained Neural Networks (Zoom, live)