SCO Seminar - Marieke Heidema and Melvyn Wildeboer, University of Groningen

Speaker 1: Melvyn Wildeboer

Title: Abstraction-Based Model Reduction of Uncertain Systems

Abstract:

Many processes related to the design and control of engineering systems depend on the availability of an accurate mathematical model describing the dynamic behaviour of the system. However, the ever-increasing complexity of modern engineering systems results in models that have such high complexity that they are no longer suitable for these purposes. The objective of model reduction is to accurately approximate models of highly-complex systems by models with a lower complexity. In particular, in this presentation an alternative perspective for performing model reduction will be introduced, in which model reduction for uncertain systems is performed in an abstraction-based setting.

Speaker 2: Marieke Heidema

Title: Learning in Memristive Neural Networks

There is an increasing interest in Artificial Neural Networks (ANNs) and implementing them in hardware. Whereas ANNs are commonly simulated in software, a hardware implementation could be more energy-efficient. In this presentation, we will make use of memristors to model ANNs in an electrical circuit. In particular, we will discuss how one can evaluate the ANN at an input using only input-output signals of the circuit. Furthermore, we will discuss how to steer the memristors so that the circuit matches a given ANN.