POLLACK PERIODICA
An International Journal for Engineering and Information Sciences
DOI: 10.1556/Pollack.4.2009.3.13
Vol. 4, No. 3, pp. 143–154 (2009)
www.akademiai.com
HU ISSN 1788–1994 © 2009 Akadémiai Kiadó, Budapest
EMBEDDED NEURAL CONTROLLERS BASED ON
SPIKING NEURON MODELS
László BAKÓ, Sándor Tihamér BRASSAI
Department of Electrical Engineering, Faculty of Technical and Human Sciences
Tirgu-Mures, Sapientia - Hungarian University of Transylvania, Calea Sighisoarei 1C
Corunca, Romania, e-mail: lbako@ms.sapientia.ro
Received 12 January 2009; accepted 20 March 2009
Abstract: This paper demonstrates, that input patterns can be encoded in the synaptic weights
by local Hebbian delay-learning of spiking neurons (SN), where, after learning, the firing time of
an output neuron reflects the distance of the evaluated pattern to its learned input pattern thus
realizing a kind of RBF behavior. Furthermore, the paper shows, that temporal spike-time coding
and Hebbian learning is a viable means for unsupervised computation in a network of SNs, as the
network is capable of clustering realistic data. Then, two versions - with and without embedded
micro-controllers - of a SNN are implemented for the aforementioned task.
Keywords: Embedded systems, Hardware/software co-design, Spiking neural networks,
FPGA
1. Introduction
The neuron models involved in Spiking Neural Networks (SNNs) are typically more
complex than in conventional rate-coded artificial neural networks and the information
passed between these neurons is expressed as temporally separated discrete events or
spikes. SNNs and Pulse-Coded Neural Networks can generate behaviors and reproduce
coding schemes closely analogous to biological neural systems. On the other hand,
these types of neural models may present important advantages in terms of digital
hardware implementation, due to the fact, that timing delay coding offers considerable
resource utilization gains and it is more robust than the analog implementation [1], [2].
An interesting proposition related to the possible applications of networks of spiking