Introduction
The course focuses on the topic of geometric deep learning that involves theoretical and practical exercise. This is a 3.0 credit course.
The course will explain the different geometric objects, such as graphs, sets, groups, grids, manifolds, etc. From the classical machine learning method to neural network approaches, covered topics include spectral theory, high-dimensional data learning, groups, geometric invariance, geometric equivariance, Fourier transform on graphs, graph convolution, geodesic, manifold, graph neural network, graph attention network, etc.
The course will be given in English.
Prerequisites
Deep learning course 236781 or a similar one.
Homework
There will be 3 homework assignments published during the semester (in intervals of 2-3 weeks), consisting of theoretical and/or programming questions. Rigorous mathematical proofs and reasoning are required for theoretical questions. Programming must be done in Python (.py files / Jupyter notebook). Submission is in pairs only.
Grade
The course grade will be based on 3 homework exercises (40%) and a final project (60%).