Enable passthrough of arguments to scipy.optimize.minimize via ClarkLDF.fit

[genedan]

Description

ClarkLDF solves for the curve parameters by using scipy.optmize.minimize:

chainladder-python/chainladder/development/clark.py

Lines 186 to 188 in 042818a

	idx_params.append(
	minimize(fun=solver, x0=x0.flatten(), bounds=bounds).x.reshape(1, 1, 1, -1)
	)

I have received a surprising amount of questions at work concerning optimization algorithms, such as which one I used, how many iterations it took, what the starting position was, etc., and then being requested to try out different variations on these attributes and then documenting them.

scipy.optimize.minimize offers a variety of algorithms and parameters, and also has the ability to return iteration-by-iteration results (something I have been asked to provide in the past) by controlling the arguments.

By allowing the passthrough of arguments in ClarkLDF.fit, we can allow users to answer such questions and accommodate these requests.

Is your feature request aligned with the scope of the package?

• Yes, absolutely!
• No, but it's still worth discussing.
• N/A (this request is not a codebase enhancement).

Describe the solution you'd like, or your current workaround.

Something like:

ClarkLDF.fit(self, X, y=None, sample_weight=None, *args, **kwargs)

Where *args, **kwargs can be passed through to scipy.optimize.minimize

Do you have any additional supporting notes?

No response

[henrydingliu]

you probably want to put the arguments inside the clarkldf object rather than the fit function