scaling of a biophysical neocortical attractor model using parallel neuron on the blue gene /p

2
POSTER PRESENTATION Open Access Scaling of a biophysical neocortical attractor model using Parallel NEURON on the Blue Gene /P David Silverstein 1,2* , Anders Lansner 1,2 From Twentieth Annual Computational Neuroscience Meeting: CNS*2011 Stockholm, Sweden. 23-28 July 2011 This work entails scaling a biophysical model of the neocortex using parallel NEURON [1] while running on a Blue Gene / P in virtual node mode. Previous scaling experiments have been done with the SPLIT simulator on the Blue Gene / L with a similar neocortical model [2]. We chose a biophysical model of medium complex- ity based on the Hodgkin-Huxley formalism because this provides the capability of exploring the effects of psychotropic drugs as well as the oscillatory effects of cortical microcircuits and globally correlated network activity. Neocortical simulations were performed to determine both strong (fixed network size, increasing cores) and weak (increasing network size, fixed load per core) scaling with two variations of a square necortical patch of hypercolumns and internal minicolumns. The first variation consists of minicolumns with 20 layer 2/3 pyramidal cells, 2 basket cells and 5 layer 4 pyramidal cells and has orthogonally stored memory patterns, encoded with long-range excitatory connections between individual minicolumns across hypercolumns. The sec- ond variation has an additional 2 regular spiking non- pyramidal interneurons per minicolumn and instead uses sparse, randomly overlapping memory patterns encoded with both excitatory and inhibitory long-range * Correspondence: [email protected] 1 Department of Computational Biology, Royal Institute of Technology, Stockholm, Sweden Full list of author information is available at the end of the article Figure 1 Strong and weak scaling results from simulations of the first variation of the neocortical model. The blue lines represent combined initialization and simulation time and the black lines also include writing spiking output. A. Strong scaling of a 16x16 hypercolumn neocortical patch with 128 minicolumns. B. Weak scaling of 4x4, 8x8, 4x8 and 16x16 hypercolumn patches all with 128 minicolumns. Each core computed 2 minicolumns. Silverstein and Lansner BMC Neuroscience 2011, 12(Suppl 1):P191 http://www.biomedcentral.com/1471-2202/12/S1/P191 © 2011 Silverstein and Lansner; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Upload: david-silverstein

Post on 06-Jul-2016

217 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: Scaling of a biophysical neocortical attractor model using Parallel NEURON on the Blue Gene /P

POSTER PRESENTATION Open Access

Scaling of a biophysical neocortical attractormodel using Parallel NEURON on the Blue Gene /PDavid Silverstein1,2*, Anders Lansner1,2

From Twentieth Annual Computational Neuroscience Meeting: CNS*2011Stockholm, Sweden. 23-28 July 2011

This work entails scaling a biophysical model of theneocortex using parallel NEURON [1] while running ona Blue Gene / P in virtual node mode. Previous scalingexperiments have been done with the SPLIT simulatoron the Blue Gene / L with a similar neocortical model[2]. We chose a biophysical model of medium complex-ity based on the Hodgkin-Huxley formalism becausethis provides the capability of exploring the effects ofpsychotropic drugs as well as the oscillatory effects ofcortical microcircuits and globally correlated networkactivity. Neocortical simulations were performed todetermine both strong (fixed network size, increasing

cores) and weak (increasing network size, fixed load percore) scaling with two variations of a square necorticalpatch of hypercolumns and internal minicolumns. Thefirst variation consists of minicolumns with 20 layer 2/3pyramidal cells, 2 basket cells and 5 layer 4 pyramidalcells and has orthogonally stored memory patterns,encoded with long-range excitatory connections betweenindividual minicolumns across hypercolumns. The sec-ond variation has an additional 2 regular spiking non-pyramidal interneurons per minicolumn and insteaduses sparse, randomly overlapping memory patternsencoded with both excitatory and inhibitory long-range

* Correspondence: [email protected] of Computational Biology, Royal Institute of Technology,Stockholm, SwedenFull list of author information is available at the end of the article

Figure 1 Strong and weak scaling results from simulations of the first variation of the neocortical model. The blue lines represent combinedinitialization and simulation time and the black lines also include writing spiking output. A. Strong scaling of a 16x16 hypercolumn neocorticalpatch with 128 minicolumns. B. Weak scaling of 4x4, 8x8, 4x8 and 16x16 hypercolumn patches all with 128 minicolumns. Each core computed 2minicolumns.

Silverstein and Lansner BMC Neuroscience 2011, 12(Suppl 1):P191http://www.biomedcentral.com/1471-2202/12/S1/P191

© 2011 Silverstein and Lansner; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the CreativeCommons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, andreproduction in any medium, provided the original work is properly cited.

Page 2: Scaling of a biophysical neocortical attractor model using Parallel NEURON on the Blue Gene /P

connections between hypercolumns. Simulations wereperformed with both single patches of increasing areaand cascades of multiple patches with feed-forward andfeed-backward projections.Individual simulations consisted of stimulation and

completion of a single memory pattern within 1 secondof cortical activity. Preliminary results show near linearspeedups of the computational part of the simulation,but degradation of file I/O performance as the numberof cores increase. Since each core writes out spikingactivity after the simulation, the performance declinemay be due to the ratio of core to I/O nodes and thelarge number of output files. With this performanceanalysis, further work will include measuring and scalingmemory storage capacity with the described second var-iation of the biophysical neocortical model.

Author details1Department of Computational Biology, Royal Institute of Technology,Stockholm, Sweden. 2Stockholm Brain Institute, Stockholm, Sweden.

Published: 18 July 2011

References1. Carnevale NT, Hines ML: The NEURON Book. Cambridge University Press;

2006.2. Djurfeldt M, Lundqvist M, Johansson C, Rehn M, Ekeberg Ö, Lansner A:

Brain-scale simulation of the neocortex on the IBM Blue Gene/Lsupercomputer. IBM Journal of Research and Development; 2008:52:31.

doi:10.1186/1471-2202-12-S1-P191Cite this article as: Silverstein and Lansner: Scaling of a biophysicalneocortical attractor model using Parallel NEURON on the Blue Gene /P.BMC Neuroscience 2011 12(Suppl 1):P191.

Submit your next manuscript to BioMed Centraland take full advantage of:

• Convenient online submission

• Thorough peer review

• No space constraints or color figure charges

• Immediate publication on acceptance

• Inclusion in PubMed, CAS, Scopus and Google Scholar

• Research which is freely available for redistribution

Submit your manuscript at www.biomedcentral.com/submit

Silverstein and Lansner BMC Neuroscience 2011, 12(Suppl 1):P191http://www.biomedcentral.com/1471-2202/12/S1/P191

Page 2 of 2