back to search

Identification of Artificial Neural Networks: from the Analysis of one Neuron to Deep Neural Networks

MA5929Elective Modules6 ECTSEnglishUnregelmäßigDepartment Mathematics
AI-edited module sheet. Based on the TUMonline module description, edited for readability.Original in TUMonline

What it is about

You will learn mathematical methods for the analysis and identification of feed-forward neural networks — from individual neurons to flat networks up to deep networks. The focus is on linear algebra, probability (in particular concentration inequalities) and optimization to determine weights and activation functions efficiently and robustly.

What you will be able to do

  • Application of linear algebra to network identification
  • Use of concentration inequalities in probabilistic analyses
  • Use of non-linear optimization methods to determine weights
  • Understanding active and passive sampling strategies
  • Identification and analysis of underparameterized, overparameterized and entangled weights

What the module consists of

  • LecturePresentation of mathematical theory and proof techniques; introduction of numerical experiments
  • Online Streamingweekly live format for attending lectures; recordings available

Teaching method

  • Weekly Online StreamingLive lectures and access to recordings
  • Handwritten representation with tablet (PDF)precise and traceable presentation of mathematical derivations
  • Slidespresentation and explanation of numerical experiments
No dates in the current semester
There are no course dates for this module this semester, or they haven't been matched yet.

Module ratings

No ratings for this module yet.

Rate this module

Only fill in the categories you can judge – for each one, either stars and text together or nothing at all.

Lecture
Tutorial
Exam

Reviews are automatically checked before they are published.

Official page in TUMonline · Details are not binding.