Parameter class

[hammannr]

This adds the Parameter and Parameters classes, which simplify the handling of our parameters. Closes #13 and #18.

Some example code to play around with the Parameter & Parameters classes:
Simple examples:

from alea.parameters import Parameter, Parameters

mu = Parameter(name="mu", nominal_value=1, fittable=True, uncertainty=0.1)

>>> mu
alea.parameters.Parameter(name=mu, nominal_value=1, fittable=True, _uncertainty=0.1)

>>> mu.uncertainty
0.1

>>> mu.uncertainty = .5
>>> mu.uncertainty
0.5

>>> parameters = Parameters()
>>> parameters.add_parameter(mu)
>>> parameters
alea.parameters.Parameters(mu)

load from config:

parameter_definition = {
    "wimp_mass": {
        "nominal_value": 50,
        "fittable": False,
    },
    "livetime_0": {
        "nominal_value": 0.2,
        "fittable": False,
        "description": "Livetime of SR0 in years",
    },
    "livetime_1": {
        "nominal_value": 1.,
        "fittable": False,
        "description": "Livetime of SR1 in years",
    },
    "signal_rate_multiplier": {
        "nominal_value": 1.,
        "ptype": "rate",
        "fittable": True,
        "fit_limits": (0, None),
    },
    "er_rate_multiplier": {
        "nominal_value": 1.,
        "ptype": "rate",
        "uncertainty": 0.2,
        "relative_uncertainty": True,
        "fittable": True,
        "fit_limits": (0, None),
        "fit_guess": 1.,
    },
    "par1": {
        "nominal_value": 0,
        "ptype": "shape",
        "uncertainty": "scipy.stats.uniform(loc=-1, scale=2)",
        "relative_uncertainty": False,
        "fittable": True,
        "blueice_anchors": [-1, 0, 1],
        "fit_limits": (-1, 1),
        "description": "ER PCA shape parameter 1",
    },
}

>>> parameters = Parameters.from_config(parameter_definition)
>>> parameters.wimp_mass.nominal_value = 100

>>> parameters()
{'wimp_mass': 100,
 'livetime_0': 0.2,
 'livetime_1': 1.0,
 'signal_rate_multiplier': 1.0,
 'er_rate_multiplier': 1.0,
 'par1': 0}

>>> parameters(signal_rate_multiplier=0.)
{'wimp_mass': 100,
 'livetime_0': 0.2,
 'livetime_1': 1.0,
 'signal_rate_multiplier': 0.0,
 'er_rate_multiplier': 1.0,
 'par1': 0}

>>> parameters.fittable
['signal_rate_multiplier', 'er_rate_multiplier', 'par1']

An updated example for the Gaussian model:

from alea.examples.gaussian_model import GaussianModel
import numpy as np
import matplotlib.pyplot as plt


model = GaussianModel()

# further define the parameters
model.parameters.sigma.fittable = False
model.parameters.sigma.nominal_value = .1
model.parameters.mu.nominal_value = 1.
model.parameters.mu.fit_limits = (-5, 5)

# toy generation
data = model.generate_data()
model.data = data

# Fit the data
best_fit, ll_val = model.fit(verbose=True)

[kdund]

Thanks @hammannr looks good! Added a nominal values property + class to update parameter values, I think this is useful to always pass all args to likelihood etc. that way (we can add limit checking next pass I guess)

[hammannr]

@kdund if I understand correctly, your get_parameters_to_call method is essentially the same as the __call__ method. And I think we could do the clean version that you implemented in ll and generate_args also with the call one since it will set the default if None is parsed. I just missed doing this 😅 ..
Did I miss the point of get_parameters_to_call ? Otherwise, I'd revert to using the call function and write parameters = self.parameters(mu=mu, sigma=sigma) in the ll and generate_data methods.

[kdund]

oops, sorry, @hammannr didn't realise I was duplicating. I do think it is neater to clean away Nones in the parameter call. feel free to remove what does not add anything

[hammannr]

@kdund I now went back to the call method, it should have all the functionality (in particular, if you parse Nones as kwargs they are just dropped).
Also, I took your nice idea of a method to check whether a value is within limit and implemented it already in the Parameter class as I think it already makes sense there. I also added an exception in case one tries to fix a value outside the fit params (I guess this was your intent, right?).
Let me know whether you think those changes fit your intended purpose! 😊

[kdund]

I would still have a method for it since that will be useful to put for the generate_data function

…

On Tue, Jun 20, 2023, 18:16 Robert Hammann ***@***.***> wrote: @kdund <https://github.com/kdund> I now went back to the call method, it should have all the functionality (in particular, if you parse Nones as kwargs they are just dropped). Also, I took your nice idea of a method to check whether a value is within limit and implemented it already in the Parameter class as I think it already makes sense there. I also added an exception in case one tries to fix a value outside the fit params (I guess this was your intent, right?). Let me know whether you think those changes fit your intended purpose! 😊 — Reply to this email directly, view it on GitHub <#19 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEAPFKPYFRG7Q5KF7AKT2VLXMIOLVANCNFSM6AAAAAAZNWCC2I> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[hammannr]

Also solves #6 (forgot earlier).