Neuromorphic Computing and Network Dynamics

At a glance | Challenges | Solutions | Contact | Research Groups

At a glance

Computers modelled on the human brain have the potential to be much more energy-efficient than conventional computers. In the topic “Network Dynamics and Neuromorphic Computing”, Jülich experts study the brain and its network dynamics. The aim is to understand the fundamental computing principles of the brain right down to the resolution level of individual neurons and synapses and to then use this knowledge for artificial systems.

Based on these insights, Jülich researchers are developing network simulations and mathematical models that are incorporated into the design of neuromorphic computer technologies. This also leads to ideas for memristive components, which are a promising aspect of neuromorphic structures.

Challenges

Around 15 % of the world’s electrical energy is consumed by IT applications – a trend that is on the rise. This is why it is more essential than ever to develop smart and energy-efficient computing systems.

The brain’s roughly 86 billion neurons and over 100 trillion synaptic connections form a natural neural network that dynamically adapts to new tasks and works very efficiently. It is the subject of intense research. The challenges of simulating this complex system in neuromorphic computing range from the development of complex network structures and algorithms to the realization of completely new materials and components.

Solutions

In order to understand the highly complex human brain and its capabilities, Jülich researchers are investigating network dynamics and their application in neuromorphic computing.

One focus of the researchers is the modelling and simulation of neural networks. Anatomical and physiological data are translated into mathematical models that depict the natural densities and structures of these networks. The development of complex simulation methods and the use of efficient software solutions are key components of this work, with the researchers particularly taking into account the requirements of the scientific community.

At the same time, Jülich researchers are developing analysis methods to reveal coordinated neural activities and patterns in extensive data sets. These findings are used to improve the network models and support research into cognitive functions and learning processes.

Furthermore, research at Jülich focuses on the transfer of these biologically inspired models and principles to neuromorphic computing. This will in future lead to new kinds of technological systems that combine data storage and data processing in an innovative way.

Neuromorphic computing is useful for numerous applications. Potential use cases range from autonomous driving and learning industrial control systems to the construction of intelligent and self-powered implants.

Jülich experts are developing ideas for memristive systems, which play a key role in neuromorphic computing. Memristors are electronic components that change their resistance depending on the charge that has passed through them and thus have a memory effect.

Jülich researchers are developing corresponding chips whose components transmit and process information like synapses in the brain. These innovative chips, which can be miniaturized down to the nanoscale, can work together with conventional microelectronics. To this end, Jülich scientists are designing detailed circuit models that describe the behaviour of memristors in specific application scenarios, such as in memories or processors. Machine learning and artificial intelligence in particular could benefit from such computer systems in future.

Comprehensive models of possible memristor designs are inspiring experts in materials research and nanotechnology. They are using the theoretical results to optimize and scale the physical properties of these components and their production for use in neuromorphic systems. In addition, the Jülich Neuromorphic Computing Alliance (JUNCA) brings together scientists from the field of neuromorphic computing. The resulting exchange of knowledge and data helps to accelerate research.