Multicore asynchronous simulation of spiking neural networks on the grid

Dynamics analysis studies of spiking neural network behavior entails the computation of a large number of simulation scenarios. Moreover, when the simulated neural micro-circuits are fairly large, the use of multiple cores in a simulation tends to be beneficial. In this paper we focus on the paralle...

Full description

Saved in:
Bibliographic Details
Main Authors: Joldos, M., Muntean, I. L.
Format: Conference Proceeding
Language:English
Subjects:
Asynchronous Simulation
Biological system modeling
Computational modeling
Globus Online Client
Globus Toolkit
Grid computing
Instruction sets
Integrated circuit modeling
Multicore processing
Neural microtechnology
Neurons
Spiking Neural Networks
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Dynamics analysis studies of spiking neural network behavior entails the computation of a large number of simulation scenarios. Moreover, when the simulated neural micro-circuits are fairly large, the use of multiple cores in a simulation tends to be beneficial. In this paper we focus on the parallelization of an asynchronous simulation strategy with OpenMP threads and on the management of the neural simulations on compute grids. The first results show a parallel efficiency in the ranges of 0.31-0.98, with hyper-threading disabled or enabled, resp. By using the grid as underlying middleware for performing the simulations, we could generate and manage hundreds of simulation scenarios in an easy way.
ISSN:2068-1038
DOI:10.1109/RoEduNet.2013.6511739