Please notice: There are no prerequisites for Master of Computer
Science students for this module in Winter 2019.
The lecture is organized as 2hrs lecture plus 2 hrs exercises per week.
The lecture starts on: Monday, 7 Oct 2019, 12ct, Lecture Hall 5+6, Endenicher Allee 19c.
This lecture is part of the intelligent systems track of the master
programme "Computer Science".
The lecture gives an overview over the most important technical
neural networks and neural paradigms.
The following topics will be explained in detail: Perceptron, multi-layer perceptron (MLP), radial-basis function nets (RBF), Hopfield nets, self organizing feature maps (SOMS, Kohonen), adaptive resonance theory (ART), learning vector quantization, recurrent networks, back-propagation of error, reinforcement learning, Q-learning, support vector machines (SVM), Neocognitron, Convolutional Neural Networks, Deep Learning.
In addition exemplary applications of neural nets will be presented and discussed: function approximation, prediction, quality control, image processing, speech processing, action planning, control of technical processes and robots.
Implementation of neural networks in hardware and software: tools, simulators, analog and digital neural hardware.
The exercises are arranged to intensify the work with the
research topics presented in the lecture. You will get weekly
paper-and-pencil assignments that are designed to be worked on in two
person groups and completed within one week. Your results of the
assignments shall be presented and discussed during the exercise group
to practice and improve your oral presentation skills. The paper and
pencil assignments are accompanied by small programming tasks to be
completed using individually implemented programms and stat of the art
To be admitted to the exam, you need a minimum of 50% of the points from the paper and pencil exercises and two successful presentations of your solutions within the exercise group.