problems using the Network Builder

The basics of how to develop, test, and use models.
tseredynski

Re: problems using the Network Builder

Post by tseredynski »

Had really hoped that was going to work... the following is the terminal output from trying the code you posted:

Code: Select all

cvode.use_local_dt(1)
        0 
I assume the 0 means that it failed to change to local dt? Looking at the check boxes in "Numerical Number Selection" still shows CVODE being active, and not Local Step.

Just to check, you had quoted NEURON version 7.3 when you were asking what version I was using. Is there a newer version for Windows I should be using? I only saw 7.2 in the download section..

EDIT: Are you able to compile that code I sent you and successfully change to local dt using cvode.use_local_dt(1)?
ted
Site Admin
Posts: 6299
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: problems using the Network Builder

Post by ted »

Mystery revealed: the bug, if there was one, lay in my failure to recognize from the onset the significance of the fact that your model consists entirely of artificial spiking cells. This explains everything.

"'current model type' is set to <*no model*>" The term "current model type" reports the nature of the system equations. Possible types include ODEs (ordinary DEs), DAEs (mix of algebraic and ODEs), and "no model" (no DEs). In recent versions of NEURON, artificial spiking cells are implemented as ARTIFICIAL_CELLs. Quoting from the comment at the start of intfire1.mod, such a model "not only has a NET_RECEIVE block and does NOT have a BREAKPOINT but is not associated with a section location or numerical integrator." So there is nothing that needs to be integrated. Each cell is updated only when it receives an input event, and there is no need for local time steps. If you want to use local time step integration, you must have at least one model cell that involves at least one section.

Years back, there was no ARTIFICIAL_CELL keyword in NMODL, so artificial spiking cells had to be implemented as POINT_PROCESSes. This meant that a network of such cells also had to have a "host" section to which the POINT_PROCESSes would be attached, and this is why local variable time step could be invoked in a network that (seemed) to consist entirely of artificial spiking cells.
tseredynski

Re: problems using the Network Builder

Post by tseredynski »

Wow, thank you so much ted. I really appreciate you continuing to help figure this out, use of this program at my university is new and leaves me with few avenues to seek help. Happy to know that everything's working as it should, and moreover why it was different from what the tutorials lead me to expect.

Thanks again!
ted
Site Admin
Posts: 6299
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: problems using the Network Builder

Post by ted »

Sorry it took me so long to realize what was going on. Yours was the first question about this particular topic that I have seen; I'm surprised it took so long for somebody to ask. I'll have to revise that tutorial..
Post Reply