Broadcast: Events

Neuromorphic computing is an emerging technology that employs silicon neurons to solve computational problems. Neuromorphic hardware systems are now becoming available in research and industry. The big challenge to the field is to design neuromorphic “algorithms” (i.e. networks and learning rules) that address real computing tasks.   

I will present a recent study in which we developed a proof of concept for neuromorphic pattern recognition in multivariate data [1]. The network we designed runs on the Spikey hardware system, which operates at a 10,000-fold speedup compared to biological neurons [2]. This study sheds light on the challenges that the design of neuromorphic algorithms involves, but also points out the potential provided by brain-like parallelism. 

I will finish by giving an outlook on our planned work and on yet unexploited features that this novel computing paradigm offers.

[1] Schmuker et al., 2014. A neuromorphic network for generic multivariate data classification. PNAS 111(6):2081–86. http://dx.doi.org/10.1073/pnas.1303053111

[2] Pfeil et al., 2013. Six networks on a universal neuromorphic computing substrate. Front. Neurosci 7:11. http://dx.doi.org/10.3389/fnins.2013.00011

By: Luke Scott
Last updated: Friday, 17 October 2014