[Feature request]: Incorporate `gempyor.distributions` Into `gempyor.inference_parameter.InferenceParameters`

[TimothyWillard]

Label

gempyor, inference

Priority Label

high priority

Is your feature request related to a problem? Please describe.

1. There is a typo at

flepiMoP/flepimop/gempyor_pkg/src/gempyor/inference_parameter.py

Line 104 in fda910d

pdist=istribution_from_confuse_config(

and

flepiMoP/flepimop/gempyor_pkg/src/gempyor/inference_parameter.py

Line 118 in fda910d

pdist=istribution_from_confuse_config(

2. The InferenceParameters class assumes that lower/upper bounds for modifiers are given via a/b, which limits them to truncated normal distributions.

Is your feature request related to a new application, scenario round, pathogen? Please describe.

No response

Describe the solution you'd like

1. Correct the typo.
2. The DistributionABC needs to be extended to expose a support property that can be used to obtain the lower/upper bounds as a tuple. Then this needs to be incorporated into the InferenceParameters class to replace the lines that look like:

flepiMoP/flepimop/gempyor_pkg/src/gempyor/inference_parameter.py

Lines 63 to 64 in fda910d

	lb=parameter_config["value"]["a"].get(float),
	ub=parameter_config["value"]["b"].get(float),

[pearsonca]

I like the 2 solution; can we tell if lb / ub are happy to be infinite or must they be like MAX_FLOAT or w/e?

[TimothyWillard]

I like the 2 solution; can we tell if lb / ub are happy to be infinite or must they be like MAX_FLOAT or w/e?

Numpy's inf should be okay. The lower/upper bound check occurs here:

flepiMoP/flepimop/gempyor_pkg/src/gempyor/inference_parameter.py

Lines 225 to 232 in fda910d

	def hit_lbs(self, proposal) -> np.ndarray:
	return np.array((proposal < self.lbs))
	def hit_ubs(self, proposal) -> np.ndarray:
	"""
	boolean vector of True if the parameter is bigger than the upper bound and False if not
	"""
	return np.array((proposal > self.ubs))

Which should support comparison with np.inf.

Edit: These methods should also be refactored to have updated type hints and take advantage of np.greater/less_equal.