Open tclose opened 9 years ago
I sort of feel that spike buffers are an implementation detail. They are sometimes needed if you're trying to handle multiple non-linear synaptic currents with a single mechanism, but I think the cleaner way would be to just have multiple mechanisms, one per synapse.
I agree that we don't want to have to specify the buffer itself but could you sometimes have nonlinear interactions between spikes on the same stream that you would want to model and therefore need to model the response to each spike separately?
On Fri, 29 May 2015 at 9:11 pm Andrew Davison notifications@github.com wrote:
I sort of feel that spike buffers are an implementation detail. They are sometimes needed if you're trying to handle multiple non-linear synaptic currents with a single mechanism, but I think the cleaner way would be to just have multiple mechanisms, one per synapse.
—
Reply to this email directly or view it on GitHub https://github.com/INCF/nineml/issues/103#issuecomment-106784731.
In the process of importing the PyNN standard models NMODL files into 9ML I came across explicit spike buffers and realised that I don't think we have a way of representing them in 9ML, e.g.
While in this case there should be a way around them I am not sure there is in general. Is this something we should look at, and if so does anyone have any suggestions, or have I missed something very obvious here?
Perhaps we could have some sort of "ReduceStateVariable", which is actually the combination of variable per event (with a time limit). But this is very hacky...