Model of Visual Attention CLAMVis MIIND Modelling the Population Level and Beyond D. G. Harrison 1 M. de Kamps 1 School of Computing University of Leeds 14 th March, 2012 M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis MIIND Funding for the Neurosciences Three drivers: Health related Brain-inspired technology Understanding brain/mi(i)nd M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis MIIND Spatial Structure M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis MIIND Local Circuit M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis MIIND M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis MIIND Circuits M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis CLAMVis XML MIIND Outline Model of Visual Attention 1 CLAMVis 2 CLAMVis XML Structure MIIND 3 M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis CLAMVis XML MIIND CLAMVis Top Level Elements M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis CLAMVis XML MIIND Layer Descriptions and Networks M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis CLAMVis XML MIIND Circuit Descriptions M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis CLAMVis XML MIIND Dynamic Networks M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis CLAMVis XML MIIND Simulations M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis CLAMVis XML MIIND Is This Neuroscience? M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis MIIND Modelling Populations: MIIND Sophisticated Population Dynamics Wilson-Cowan Dynamics Population Density method Describes large 3 × 10 20 20 0.25 f (Hz) f (Hz) ρ populations of 18 18 16 16 0.2 leaky-integrate-and-fire 14 14 neurons 12 12 0.15 10 10 Fokker-Planck but better; 8 8 0.1 no diffusion limit 6 6 ρ ( v ) dv : fraction of 4 4 0.05 2 2 population with membrane -3 -3 -3 × 10 × 10 × 10 0 0 0 0 20 40 60 80 100 120 140 160 180 200 12 13 14 15 16 17 18 19 20 21 22 0 5 10 15 20 25 potential in [ v , v + dv ) t (s) (V) V (V) µ Balance excitation-inhibition M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis MIIND Short Demo M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis MIIND Conclusions Many ’cognitive’ models use stereotypic coding patterns which can easily be captured in XML The population is a natural bridge between ’basic’ neuroscience and more higher level models The population level scales well, although it may not be appropriate in every situation Much replication effort can be avoided by extending the NeuroML, NineML domains Very much in the interest of the neurosciences, given the drivers for funding M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis MIIND Status MIIND: public; XML available (slightly brittle). Documentation, patchy, but improving: http://miind.sf.net Soon: generic 1D neural model solver (not just LIF: QIF as well) Aim: Generic 2D population solver Izhikevich adaptive exponential synapses Working on cloud implementation, web interface, tutorial ClamVis: Dave’s project. Not really public but check out: http://stacker.me.uk/~daveh/NetSimDocs/ projectxmlformat.html M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Model of Visual Attention CLAMVis MIIND Acknowledgement Dave Harrison: modelling, XML stuff Robert Cannon Padraig Gleesson, Angus Silver M. de Kamps, D. G. Harrison Modelling the Population Level and Beyond
Recommend
More recommend