`bounds` cannot be used together with `method=Brent` in latest version of scipy (>= v1.10.1)

[thupchnsky]

SciPy (>= v1.10.1) will complain about this line

autodp/autodp/converter.py

Line 95 in 5fad5e1

results = minimize_scalar(fun, method='Brent', bracket=(1, 2), bounds=[1, 100000])

because it now does not support using Brent when a bound is given (scipy source)

if bounds is not None and meth in {'brent', 'golden'}:
    message = f"Use of `bounds` is incompatible with 'method={method}'."
    raise ValueError(message)

Can switch to method='Bounded' to bypass this issue.

[jeremy43]

Thank you for pointing out this issue and the suggestion! We will run a test on switching method to `bounded'. Will update this part if it passes all test cases.

[philipjk]

In my case, even using 'Bounded' gives errors as it expects both bounds to be finite, whereas alpha_max (in rdp_to_approxdp) defaults np.inf. In particular: ValueError: Optimization bounds must be finite scalars.
As per the documentation of minimize_scalar, both bounds need to be finite. I tried with np.finfo(np.float64) (the largest representable float64) but it seems like it interprets it as infinity. Right now, I'm just giving it a large value manually to bypass the issue, like 1e3 or something.

[yuxiangw]

This is an interesting case. "bounded" were used when there is a natural bounded range, and I implemented a doubling trick to get the right bound for some of the numerical computation tasks in autodp. "brent" was tested on other cases when a range is not available beforehand. It seems that we can still use "brent" in those cases, it's just the dummy bounded range input should be removed.

@jeremy43 can we test it out? Also, there are many cases in the repo where "minimize_scalar" is used. There are a lot of repetition right now. I think we should categorize how many each type of application of "minimize_scalr" there are and modularize them. This way we do not have a lot of repeated code to maintain.

Will keep the issue open for now until it's resolved.

[khangtran2020]

I have the same problem using autodp v.2.0. And also I can not install autodp with version 2.1.0b0 through pip.