Mathematics of Machine Learning (and AI)
The theoretical analysis of deep neural networks has made significant progress in the last years. The mathematical language and concepts which provide the basis for this progress can be found in the area of functional analysis. This course provides a thorough introduction to functional analysis and relates these abstract concepts to dep neural networks. Topics are vector spaces, metric spaces, scalar products, Banach- and Hilbert spaces, Sobolev spaces and Hilbert spaces, approximation theory for neural networks, basic theory of bounded linear operators, and reproducing kernel Hilbert spaces.
The students know and understand the basics of functional analysis and can apply these competences in the area of machine learning and artifical intelligence. They understand the theoretical properties of machine learning approaches. They are able to connect practical experience on the design an behavior of neural networks to their theoretical properties.
The course is based on a mixture of asynchronous online material (videos and script) for self study and in person lectures. Guided problem solving classes will help the students to deepen their understanding and to improve their self study abilities. The students will work on regular assignments for practicing. Each student will also give a presentation on a selected topic of the course.
At the end of the course, students will do a written exam.
- Wendland, Holger. Scattered Data Approximation: Scattered Data Approximation. Cambridge University Press, 2004.
- Master of Science in Artificial Intelligence, Lecture, Elective, 2nd year
- Master of Science in Artificial Intelligence, Lecture, Elective, 1st year
- PhD programme of the Faculty of Informatics, Lecture, Elective, 1st year (4 ECTS)
- Calculus, Hormann K., Chang Q., Fuda C., Georgiou A., Laneve L., SA 2022-2023
- Linear Algebra, Pivkin I., Scarpone M., Tufano R., SP 2022