Support of observable-dependent sigmas

[dweindl]

Currently it is not possibly to have state-dependent sigmas. Is it only about changing respective function signatures, or will that require further adaptations?

[FFroehlich]

Would also require a couple of additional derivatives to be taken into account. On top of my head I think its only dJydx that needs to account for dJydsigma*dsigmadx, but it might be more.

[paulstapor]

I would argue for implementing them as observable-dependent, rather than stat-dependent. That should do the job as well and one does not have to refactor observables from states...

[yannikschaelte]

Yes, @elbaraim confirmed that observable dependency is what she needs in her case.

[elbaraim]

Yes, @elbaraim confirmed that observable dependency is what she needs in her case.

In this specific case the observables are a one-to-one mapping from the model states

[JanHasenauer]

if think that observable dependence will cover most cases. Currently, I would not bee aware of anything there we need state dependence.
If observable dependence is much easier, I would go for this.

[paulflang]

I am also running into this issue atm. What I would need is observable dependence.