1. Deutsch
  2. English

Module Description

Mathematics of Neural Networks

Courses:

TitleTypeHrs/WeekPeriod
Mathematics of Neural NetworksLecture2Winter Semester
Mathematics of Neural NetworksRecitation Section (small)2Winter Semester

Module Responsibility:

Dr. Jens-Peter Zemke

Admission Requirements:

None

Recommended Previous Knowledge:

  1. Mathematics I-III
  2. Numerical Mathematics 1/ Numerics
  3. Programming skills, preferably in Python

Educational Objectives:

Professional Competence

Theoretical Knowledge

Students are able to name, state and classify state-of-the-art neural networks and their corresponding mathematical basics. They can assess the difficulties of different neural networks.

Capabilities

Students are able to implement, understand, and, tailored to the field of application, apply neural networks.

Personal Competence

Social Competence

Students can

Autonomy

Students are able to

ECTS-Credit Points Module:

6 ECTS

Examination:

Oral exam

Workload in Hours:

Independent Study Time: 124, Study Time in Lecture: 56


Course: Mathematics of Neural Networks

Lecturer:

Jens-Peter Zemke

Language:

German & English

Period:

Winter Semester

Content:

  1. Basics: analogy; layout of neural nets, universal approximation, NP-completeness
  2. Feedforward nets: backpropagation, variants of Stochastistic Gradients
  3. Deep Learning: problems and solution strategies
  4. Deep Belief Networks: energy based models, Contrastive Divergence
  5. CNN: idea, layout, FFT and Winograds algorithms, implementation details
  6. RNN: idea, dynamical systems, training, LSTM
  7. ResNN: idea, relation to neural ODEs
  8. Standard libraries: Tensorflow, Keras, PyTorch
  9. Recent trends

Literature:

  1. Skript
  2. Online-Werke:

ECTS-Credit Points Course:

6 ECTS