Alexander Held

I believe so, @will-cern can presumably confirm.

I need to have a more detailed look, am I understanding correctly that this effectively defines new uncorrelated parameters? When reading this issue I at first expected multi-dimensional Gaussian constraint...

Indeed this can also be solved externally without too much effort (and for some specific cases that might be the best way). I think an "and" combination would provide flexibility...

I gave this a try, here's an implementation: https://gist.github.com/alexander-held/b020445d089c3ef79abe93d63543c5e7. Seems to be working, but I have not tested it extensively. There are a few to-do items and some cleanup needed....

> Importantly, I've addressed an oversight, where the covariance matrix (hess_inv) was not treated correctly during fits with fixed parameters -- it's right to zero out the uncertainties, but then...

What's the central value in this case? I'm guessing the issue might be that the lower default POI bound of 0 is close to the best-fit value and therefore the...

If it is just the histosys, I am not really sure, but given the big difference you could propagate this through to the post-fit yield uncertainties, they should be comparable...

How does MINOS compare for the parameter uncertainties?

I don't have an example unfortunately, from a quick test with a higher level library I did not see the init settings get written to the xml at all for...

I am not sure whether or not it would be feasible to access `numpy` operations via `keras_core.src.ops.numpy` and if they are sufficiently complete, from a brief look I think there...