Seminar S24

High data volumes make artificial neural networks more and more indispensable in modern science. This seminar gives an intuitive introduction to the basic concepts of neural networks, specially adapted to physicists.

The core concepts of neural networks have their foundations in statistical mechanics, i.e. probability theory and statistics. We can for example draw parallels between learning methods of neural networks and entropy or between the cognition of neural networks and the Ising model. The seminar covers topics from simple perceptron models to deep neural networks and Boltzmann machines.

The students are expected to participate actively by giving a presentation (with documentation) of their topic and writing a simple code for a neural network.

Requirements:

Statistical mechanics: Gibbs Entropy, Probability distribution functions, Ising Model

Basic programming: E.g. Python or Matlab
Language: English or German

First meeting on Thursday, 11 April 2024 at 15:15 in P912

Link zu Zeus