Mathematical Methods of Machine Learning and Data Science

Abstract of the Lecture:

The lecture introduces basic mathematical methods required to understand both classical approaches and their connection to the ingredients of deep learning architectures: convolution and mathematical signal processing, data embedding and the impact of high dimensions, randomization and concentration of measure, measure transport, elementary Riemannian geometry and flows realized by networks.

  • Exercises:Jonathan Schwarz
  • Language: English
  • SWS: 4
  • ECTS: 8
  • Lecture Id: MM35, Spezialisierungsmodul Numerik und Optimierung
  • Registration:Registration will be announced
  • Prior Knowledge: Foundational courses on Linear Algebra and Analysis