Abstract
Emerging research areas in neuroscience are requiring simulation of large and detailed spiking neural networks. Although eventdriven methods have been recently proposed to simulate these networks, they still present some drawbacks. To obtain the advantages of an eventdriven simulation method and a traditional time-driven method, we present a hybrid method. This method efficiently simulates neural networks composed of several neural models: highly active neurons or neurons defined by very-complex model are simulated using a time-driven method whereas other neurons are simulated using an event-driven method based in lookup tables. To perform a comparative study of this hybrid method in terms of speed and accuracy, a model of the cerebellar granular layer has been simulated. The performance results showed that a hybrid simulation can provide considerable advantages when the network is composed of neurons with different characteristics.
Jesús Garrido
Associate Professor
Associate professor in Computation technology, senior researcher at the Computational Neuroscience and Neurorobotics Lab and principal investigator of the VALERIA lab of the University of Granada.
Niceto Luque
Associate Professor
Associate Professor at the Department of Computer Engineering, Automation and Robotics and Principal Investigator at the Applied Computational Neuroscience Group.
Ríchard R. Carrillo
Assistant Professor
Associate Professor at the Department of Computer Engineering, Automation and Robotics and Principal Investigator at the Applied Computational Neuroscience Group.
Eduardo Ros
Full Professor
Full professor in computer architecture, principal investigator at the Computational Neuroscience and Neurorobotics Lab and principal investigator of the VALERIA lab of the University of Granada.