TU Dresden
A joint project with Uni Leipzig DE EN

Neuromorphic Information Processing: Hardware Realization of the Modified Stochastic Synaptic Model

Mentored by Martin Bogdan, Thomas Mikolajick
at Leipzig University, Faculty for Mathematics and Computer Science, Institute for Computer Science, Neuromorphic Information Processing

Spiking Neural Networks (SNN) are the upcoming generation of artificial neural networks, especially in Neuromorphic Information Processing. Up to now, even though SNNs are comparable with classical machine learning in terms of performance, they still lack in terms of learning methods because of the comparatively slow training.

Moreover, important parts of the information processing are not yet included. One of these is the synaptic dynamics regarding the used resource neurotransmitter and its behaviors in the synaptic cleft. To overcome this, the department has evolved the Modified Stochastic Synaptic Model (MSSM). In order to speed up training, this algorithms have to be implemented in hardware. To do so, circuits for the complex MSSM have to be developed.

Work Environment

You will be working with Prof. Dr. Martin Bogdan and his research group for Neuromorphic Information Processing. You will have access to the equipment of the department. You will be supported as well by Prof. Dr.-Ing Thomas Mikolajick and Dr. Andre Heinzig as this work is assumed to be performed in close collaboration.

Prerequisites

To conduct this research, you should hold a very good university degree (MSc or an equivalent) in Computer Science with focus on Computer Engineering and Neuromorphic Information. Deeper knowledge in Computer Engineering especially regarding semiconductors is required.