69 fr extend set of available ansaetze

[paxuy]

Closes #69

The last four commits, were caused because i saw some minor mistakes and used "Undo Last Commit" in VSCode after merging.

[pep8speaks]

Hello @paxuy! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2025-01-23 09:54:25 UTC

[stroblme]

Thanks for all the work! 💪 I checked and found some discrepancy to the values provided in https://doi.org/10.1002/qute.201900070 .
We need to check what is going on there..
All circuits look good, thanks 🙏
General note: Circuit 10 deviates from the implementation because we build circuits layerwise which is not working with the proposed ansatz.

[stroblme]

Unable to fix the issue with discrepancy to the results in paper right now. Implementation of metrics looks fine and relative values behave as expected but absolute values deviate by up to 20% for expressibility and up to 50% for entangling capability to what is shown in the paper. Maybe @majafranz could have a look here?
To get this (temporarily) fixed, I adapted the tolerance to the deviation observed and removed all 3-layer expressibility tests (also to reduce the time it takes to run all tests).
Some observations:

• the entangling capability for most circuits is underestimated by an factor of ~0.5.
• higher expressive circuits (lower KL divergence) seem to be more prone to errors (up to 50% error) (numeric issues?)

[majafranz]

Thanks a lot! 🔥