possibility with constructing large-scale network

[breakwave922]

Hi,

I'm trying to scale my current project of simulating 1000-cell network to ~27,000-cell network in NEURON. I use to generate connectivity matrix file in matlab, with 1000*1000 elements, and saved as txt file that was later imported into NEURON for connecting cells. But if network is with 27,000 cells, such a file will be huge, and it's sometimes impossible to store and imported.
I'm thinking how can I avoid generating those connection matrix files in matlab, and generate connections on-line in paralleled NEURON. They are all random, but have some tricky parts: it is composed of unidirectional ITN-ITN, bidirectional ITN-ITN, unidirectional ITN-PN, reciprocal ITN-PN and gap junctions.

Anyone here can give me some references or some suggestions on what tool I can use to make it possible.

Thanks in advance.

[ted]

Use an algorithm to set up the network. It's up to you to figure out what algorithm would generate the connectivity you want.

[salvadord]

Another option that you could consider is using NetPyNE (a tool for NEURON), which allows you to define connectivity rules using a high-level declarative format -- see this post for more information: viewtopic.php?f=45&t=3670