Project code: ScaNN
Research group: Computational Intelligence and Multi-agent Games (SEN4)

Artificial neural networks are adaptive, learning systems that abstract from our functional understanding of how biological nervous systems work. However, research of real biological neurons is ever increasingly revealing that this abstraction of neural information processing is too simple. This has led to the creation of new, more detailed models of neural computation, abstracted into novel artificial networks of spiking neurons.

The SCANN project aims to develop methods for scaling up Artificial Spiking Neural Networks (ASNN) to large, complex problem domains. Scaling up neural networks is an important open problem that limits the range of applications for which artificial neural networks can be used. We believe that only when considering such challenging problem domains that the increased computational and communicational power of Artificial Spiking Neural Networks can be fully exploited.

The project involves research that should allow a network of spiking neural networks to function as a modular semi-independent agents, where we aim to exploit the computational and communicational advantages of spiking neural networks using and extending methods for traditional neural networks. The research aims to advance a flexible adaptive distributed computational framework for spiking neural networks that promises to be very generally applicable in a wide range of complex application fields. Additionally, we expect to also generate new insights into important neuroscience issues like "neural coding", and learning in real spiking neural networks.

Member
Stefan Bohte