Page 1 of 1

Cal4 mechanism: convergence, scopmath library error

Posted: Mon Jul 21, 2014 8:12 am
by benji
Hi, this question is similar to the problems discussed previously (http://www.neuron.yale.edu/phpBB/viewto ... ions#p7789 and http://www.neuron.yale.edu/phpBB/viewto ... ions#p7757), but I can't figure out how to solve this. I am running Asshad and Narayanan's model (http://senselab.med.yale.edu/modeldb/Sh ... del=150551). Each synapse includes a mechanism called cal4, basically a calcium buffer. When I run the model with certain (very strong) stimulation protocols, I get the following errror:

Code: Select all

at line 94 in file cal4.mod:
      SOLVE state METHOD sparse
Error at section location apical[5](0.5)
Convergence not achieved in maximum number of iterations
/opt/neuron/nrn/x86_64/bin/nrniv: scopmath library error
 in heterosynaptic_plasticity.hoc near line 22
 run()
      ^
        fadvance()
      advance()
    step()
  continuerun(1.5e+06)
and others
I'd really like to get around this, and I'd be happy to provide further details by email.

Re: Cal4 mechanism: convergence, scopmath library error

Posted: Mon Jul 21, 2014 1:00 pm
by ted
I wonder if it would help to include a CONSERVE statement for hc and ho at the end of the KINETIC block. I'm not entirely sure about the syntax for dealing with an array of STATEs--maybe it can be written as

Code: Select all

FROM i=0 TO Nannuli-1 {
  CONSERVE hc[i] + ho[i] = 1
}
or maybe something like this

CONSERVE i, hc, ho { hc + ho = 1 }

or maybe you'll have to write it out as

hc[0] + ho[0] = 1
hc[1] + ho[1] = 1
etc.

Re: Cal4 mechanism: convergence, scopmath library error

Posted: Mon Jul 21, 2014 4:50 pm
by hines
Please send me a zip file containing all the necessary hoc, ses, mod, py files needed to reproduce the problem and I'll try to diagnose what is going wrong with the solver.
Send to michael dot hines at yale dot edu

Re: Cal4 mechanism: convergence, scopmath library error

Posted: Tue Jul 22, 2014 4:40 am
by benji
Thanks for your replies, both of you. I've sent an email to Michael Hines with the code attached.