Improvements to netpyne-based HNN model

jonescompneurolab / hnn

The Human Neocortical Neurosolver (HNN) is a software tool that gives researchers/clinicians the ability to develop/test hypotheses on circuit mechanisms underlying EEG/MEG data.

http://hnn.brown.edu

Other

80 stars 25 forks source link

Improvements to netpyne-based HNN model #225

Open salvadord opened 4 years ago

salvadord commented 4 years ago

More efficient connectivity – currently all-to-all and most weights=0
Fix cell locations and distance-dependent connectivity (1um spacing)
Make dipole code more generic – currently hardcoded for L2pyr + L5pyr
Remove ad hoc parameter rules from code (including those to plot dipole signal)

salvadord commented 4 years ago

@blakecaldwell I fixed the segmentation fault issue (there was a missing line of code h.setpointer in compartCell.py)

However, I'm now getting a different dipole signal that the orig HNN.

Did you manage to get it working in your old hnn-core version? ie. the dipole plot was same as original?

thanks!

blakecaldwell commented 4 years ago

@salvadord I don't think I ever could make a good comparison to original HNN. When I ran hnn-core (wrapped around netpyne) with the code I submitted to netpyne, I recall that the output looked reasonable, but that was with the netpyne model.

However, maybe there was a problem in the code in the PR vs. what I was using if I missed a h.setpointer line. I'm sorry that that happened.

I recall making the changes in https://github.com/Neurosim-lab/netpyne/pull/470 in netpyne and getting it to run, but I did not test that very much.