Deep Learning
IMPRS CoNI Lecture Series
- Start: Apr 15, 2025 11:15 AM (Local Time Germany)
- End: Jul 8, 2025 12:45 PM
- Speaker: Nico Scherf
- Location: Max Planck Institute for Human Cognitive and Brain Sciences
- Room: Lecture Hall Gustav Theodor Fechner (C101)
- Host: IMPRS CoNI
- Contact: imprs-coni@cbs.mpg.de
Registration:
Doctoral researchers of IMPRS CoNI: via OpenCampus
Interested students from partner institutions of IMPRS CoNI: Please contact imprs-coni@cbs.mpg.de
Maximum number of participants: 100
Course credits: 2 ECTS CPs
Conditions: 80% attendance and and submission and positive evaluation of written assignment
Course description:
This course provides a fundamental exploration of deep learning, aiming to build a strong conceptual understanding of how and why it works. Starting with
supervised learning using shallow and deep neural networks, it covers essential optimisation topics such as loss functions, gradients, initialisation, performance
measurement, and regularisation techniques. We will explore specialised architectures such as convolutional neural networks for image processing, transformers for sequence modelling, and graph neural networks for graph-structured data. In unsupervised learning, the course explores generative models, including GANs, variational autoencoders, and diffusion models. It also introduces deep reinforcement learning and concludes with a discussion of the ethical considerations of deep learning. By focusing on core concepts, this course will help you develop intuition in the field of deep learning.
Suggested literature: https://udlbook.github.io/udlbook/