rht

Another alternative is to turn [pyDelaunay2D](https://github.com/jmespadero/pyDelaunay2D) into a Python package that can be installed via pip.

Though a `subprocess.Popen(["solara", "run", "app.py"])` would have worked too.

More context: I started with random initial $\beta$ and $\gamma$, not from a precomputed one. Because I wanted to see if I could reproduce from scratch. Is this the reason?...

What I meant by random is that, instead of ```python gamma, beta = parameter_utils.get_best_known_parameters_for_LABS_wrt_overlap_for_p(N, known_p) u, v = parameter_utils.to_basis(gamma, beta, basis='fourier') init_u, init_v = parameter_utils.extrapolate_parameters_in_fourier_basis(u, v, p, 1) ``` I...

Update: I tried starting from `parameter_utils`'s best known p at 29, and then extrapolating it, and then doing the COBYLA optimization. What I got is relatively smoother, but there are...

I was referring to this paper: https://arxiv.org/abs/2308.02342.

Thank you for being responsive. I just tried `p=30` (EDIT: this is without extrapolation at all, as you said) and then ran COBYLA on the params. I got ``` Success...

One more question: In the paper, elements of $\beta$ are always larger than that of $\gamma$, but in my plot, it is the reverse, indicating I had accidentally switched up...

I figured out the situation with the jaggy lines: turns out that the cached params coming from `qokit.parameter_utils.get_best_known_parameters_for_LABS_wrt_overlap_for_p` (the known solution) are jaggy to begin with. I made 3 types...

Update: Ignoring the jaggy line issue, I reproduced figure 3B from scratch without using cached result! :tada:  The code can be found at https://github.com/rht/QOKit/blob/main/minimal_reproducer_from_scratch_v1.py. I think the remaining unresolved...