Sam Yates
Sam Yates
Commenting now, because it is still relevant: unit tests and nsuite validation aside, we don't have a basic make-sure-the-spikes look correct test.
Renamed issue, to be specific to spike output test. Not exactly a unit test, but we can shove it under there for convenience.
Our admittedly vague proposal for adding random sources in the first instance is to model them as random processes, accessed via special function-like symbols in NMODL. This approach should work...
It will be simplest if we can restrict these random processes so that in the first instance at least, they are only used in NET_RECEIVE blocks: treating them correctly inside...
I think here, rather, we need a representation of the stochastic differential equation — we can then solve that to first order with a (forward) Euler scheme, where the noise...
In NMODL, we could introduce a term that corresponds to the Gaussian white noise term in the cited equation, that will in turn trigger an SDE interpretation. And then as...
Regarding the time step, yes, that's the proximate implementation concern. But more broadly, if we keep the mechanism descriptions 'mathematical' in NMODL and in the future, arblang, then we can...
This has become newly relevant with the prospect of new support from the in-situ pipeline project .
Documentation structure is currently in flux; we will return to this after reorg.
Historically, we haven't used `fmt` because it was an external dependency which we thought we didn't need, and besides, we'll essentially get it for free with C++20. That said, as...