Applied Machine Learning and Simulation

Motivation


For many problems such as image recognition, prediction, or classification, it is difficult to specify explicit solution algorithms. Machine Learning (ML) methods allow computers to learn how to solve such problems using examples. Due to the rapidly increasing availability of computing power (e.g. cloud, GPU computing), it is now possible to solve real application problems in this way. Examples of applications for which the group has generated AI solutions:

Multi-Level Simulation
Learning model predictive control for air conditioning of battery electric vehicles
Forward-looking financial forecasting system
flood early warning system
Autonomous recycling plant


Approach


In our Research Group we use ML- and data-based methods for the development of example-based assistance systems and ML-based control systems. However, many of these methods require a multitude of structural parameters (e.g. number of layers in neural networks). We are therefore looking for methods that do without these parameters, or for engineering methods that enable users to determine these parameters.

RG_KI_new.png

Contact: Stefan Wittek
                Peter Engel

More Information