Initialization in heterogeneous network

Moderator: wwlytton

Post Reply
abogaard
Posts: 8
Joined: Mon Feb 05, 2007 4:21 pm
Location: ann arbor, mi

Initialization in heterogeneous network

Post by abogaard » Wed Apr 16, 2008 8:12 pm

When creating a network of mixed cell composition (different HH type cells), NEURON will solve the initialization based upon the dynamics of the last cell created (last cell template called). Thus the excitability of other cells is affected, and this small 'corrective' term can be enough to cause another cell in the network to fire unprovoked.

How can I rewrite the init procedure so that it is cell specific in the network?

Thanks much

ted
Site Admin
Posts: 5566
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: Initialization in heterogeneous network

Post by ted » Thu Apr 17, 2008 11:11 am

abogaard wrote:When creating a network of mixed cell composition (different HH type cells), NEURON will solve the initialization based upon the dynamics of the last cell created (last cell template called). Thus the excitability of other cells is affected, and this small 'corrective' term can be enough to cause another cell in the network to fire unprovoked.
On first reading, I thought "Stuff and nonsense! A model implementation may suffer
from many problems that cause incorrect initialization, but this isn't one of them. Cells aren't
initialized seriatim. Their equations are part of the same system, all of which is initialized
together."

Then I realized you're probably not talking about initialization of a simulation, i.e. what
finitialize() does. Most likely you are referring to a "model setup" problem, which may be
caused by a buggy model specification in hoc.

The key to answering your question is to first isolate the source of the problem. You can
take a big step in the right direction by creating a minimal model that reproduces the
symptom--include just enough hoc code, mechanisms, cells, etc. so that the symptom
still persists. After you have done that, if the answer doesn't jump right out at you, it will
be necessary for us to reproduce the problem if we are to tell you its cause and cure--
in which case you would zip up just those hoc, ses, and mod files that are needed by the
minimal model, and send them to me
ted dot carnevale at yale dot edu

Post Reply