Document Actions

You are here: Home / Bernstein Conference / 2017 / Satellite Workshops / Topology and dynamic of neuronal networks as guidelines for memristive computing systems (Kohlstedt, Ziegler)

Topology and dynamic of neuronal networks as guidelines for memristive computing systems (Kohlstedt, Ziegler)


Biological systems are unmatched in their efficiency in performing cognitive tasks, such as pattern recognition, with extremely low power consumption. Therefore, it is no surprise that attempts have been made to develop bio-inspired electronic circuits, so-called neuromorphic systems, with the goal to reach the performance and power efficiency of biological systems. In this context, analogue very large-scale integration (VLSI) based on silicon complementary metal–oxide–semiconductor (CMOS) technology might provide advantages to software-dominated neuro-informatics in terms of power dissipation, computational time, and needed datasets. Recently, this field has gained new momentum with the advent of memristive devices (memristors). Memristive devices are non-volatile passive devices which can adapt their resistance to the previous charge flow. Due to their simple two-terminal capacitor-like device structure (metalinsulator-metal) they are highly attractive for neuromorphic computing: they allow realizing the central building-blocks in bio-inspired artificial neural networks in an elegant and efficient manner and can outperform the classical CMOS technology in terms of scalability, power consumption, and design flexibility.
While mainly concepts of cellular correlates of learning and memory formation were in the focus of interest, so far, a transition to a multidimensional network structure is necessary to emulate higher brain functionalities. Particularly, circuit models are required which are take prominent neuronal dynamical functionalities of single neurons, neural assemblies, and topological aspects of neural networks equally into account. This links the research on memristive devices and electronic circuits to complex networks. In this field a considerable number of theoretical studies investigate the synchronization of coupled neural oscillators including non-ideal (realistic) factors such as noise and signal delay. While synchronization mechanisms were mainly theoretically investigated, so far, fewer attempts have been tried to use electronic oscillator circuits. In this respect, memristive devices combined with pulse-coupled oscillators may offer novel opportunities for cognitive computing. This shall be discussed in the proposed workshop among top scientists in the fields of device physics, neuromorphic engineering, complex dynamics, and neural computation.

The workshop aims to explore the relevance of neuronal inspired network topologies and their dynamics for future memristive based neuromorphic circuits. The organizers seek to bring together the scientists from the fields of theoretical neuroscience, non-linear dynamics, and memristively based neuromorphic circuits to explore their overlap. The workshop aims to offer a platform to spark discussion towards promising and exciting future directions, such as cognitive computing and memristive brain chips.