-
Notifications
You must be signed in to change notification settings - Fork 45
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Discussion of possible types for buffers #368
Comments
I assume that the buffers you mention here are the
|
Just in case, I'm also referencing the post that summarizes a long discussion on how we might simplify the interaction between spikes and state variables, just so that the unit discussion does take into account the way the buffer might be used. Because of this, I would be in favour of dimensionless buffers where the unit is added by the convolution function depending on the target variable that will be affected (this includes the possible existence of a normalization factor. |
@heplesser NESTML is modeling language for nest and, therefore, I would appreciate nest interpretation. Current buffers, are currently fixed to be
@Silmathoron: I contributed in #330 |
@DimitriPlotnikov why are your spikes buffers in |
@Silmathoron it is a |
Yeah, but given the method you propose for the synaptic current (using I think any other choice would lead to non-general solutions where we would have to deal with special cases. Therefore, I propose spike buffers as pure spike storage which are coupled to the other variables through the |
@Silmathoron What do you mean with normalization factors? Initial values? |
We had a discussion about this internally on Tuesday and came to the conclusion that actually the shape should be dimensionless and the buffer should have the unit. There are a number of reasons why we think this is the right way to go:
@Silmathoron: I think that this is the most general solution we could have. If we did it the other way around, there would be no straight-forward way for the |
@DimitriPlotnikov normalization for alpha shapes, for example (so that the synaptic weight is actually the maximum value of the PSC). @jougs You mean that the buffer gathers the weighted inputs from different synapses that arrive on a single neuron, right? But the synaptic weight is just a factor and is interpreted in terms of a dimensioned PSC only once the model is known, so I'm not sure I understand why this would be a problem. And I didn't get the last auto-documentation part :s I must say I'm still unsure about having a same object (a spike buffer) get different units depending on the model... are you really sure there is no other way? So for example how would you handle the dimensions in the model @DimitriPlotnikov started to write in #330? |
Proposal for the discussion:
New syntax for the shapes as ODEs
|
@PTraeder Any progress? |
It is going rather slower than I expected. I have to fight the type system at every step. And sorry for the long response time, the notification was burried in MC spam. |
The problems I ran into stemmed from me wanting to adapt the type system to still make a distinction between |
Sounds reasonable. |
I support @jougs suggestion above and find the � Just one question: I am a bit confused by |
@heplesser |
Introduces Initial values block (cf. #368 for the discussion)
Intoruced typed buffers as discussed in #368
Done in #385. |
We are planning to add typing to buffers.
Current buffers are intended to carry the type
pA
.Concerning spike buffers, we would like to have feedback for types you consider probable/useful.
For example, would you like to see just a specific subset of units? All of the "native" NEST units?
Compound units such as
mV/ms
? Primitive types?Additionally, your ideas about possible syntax for these types would be welcome.
Currently buffers are defined thusly:
My proposition, in accordance with suggestions made by @heplesser, would be to simply add the respective type after the buffer name. Similarly to variable definitions.
for a buffer of type ms.
The text was updated successfully, but these errors were encountered: