Korbinian Kottmann

Actually, [this repository](https://github.com/matteoacrossi/adapt_ic-povm/) has it implemented in qiskit. The relevant file is [here](https://github.com/matteoacrossi/adapt_ic-povm/blob/main/neven.patch) with all the diffs, should be just a matter of copy-paste as far as I can tell...

Right now the build has trouble importing `classical_shadow` from pennylane even though this is branched against `dev` - any idea what is going wrong here?

Update: This version is now somewhat complete in terms of its contents and the points I am trying to make. However, it is not really polished yet. Especially, the code...

Good catch, it was still in my browser's cache and it got lost with the big merge with the master branch 👍 Btw, @KetpuntoG do you have an idea for...

The chain rule is working properly here, that _is_ the _problem_ 😄 Essentially, instead of $\partial_x f(g(x)) = f'(g(x)) \partial_x g(x) = f'(g(x)) (g(x+\pi/2) - g(x-\pi/2))/2$, I want $\partial_x f(g(x))...

That sounds very promising! The heavy lifting would be in the `shifts` argument here though, right? Or would this be a non-AD version where the user is expected to know...

For me this sounds generally like a desirable feature to have and it could substantially clean up some of the devices codes. I have to admit I still lack insight...

Hello @tgag17 , that would be great! Feel free to open a work-in-progress [WIP] pull request (PR) and ask questions. Whenever you have a working version, tag me or someone...

Hi @Miguelvs23 I think the problem are some unintended indents and many trailing whitespaces. You can check your code with `pylint pennylane/your/file.py` and it will tell you what he does...

We noticed one problem when the number of wires is just 1, then the return type always gets squeezed. In the case of one wire, we would expect the output...