# custom_model¶

astropy.modeling.custom_model(*args, fit_deriv=None, **kwargs)[source]

Create a model from a user defined function. The inputs and parameters of the model will be inferred from the arguments of the function.

This can be used either as a function or as a decorator. See below for examples of both usages.

Note

All model parameters have to be defined as keyword arguments with default values in the model function. Use None as a default argument value if you do not want to have a default value for that parameter.

Parameters
funcfunction

Function which defines the model. It should take N positional arguments where N is dimensions of the model (the number of independent variable in the model), and any number of keyword arguments (the parameters). It must return the value of the model (typically as an array, but can also be a scalar for scalar inputs). This corresponds to the evaluate method.

fit_derivfunction, optional

Function which defines the Jacobian derivative of the model. I.e., the derivative with respect to the parameters of the model. It should have the same argument signature as func, but should return a sequence where each element of the sequence is the derivative with respect to the corresponding argument. This corresponds to the fit_deriv() method.

Examples

Define a sinusoidal model function as a custom 1D model:

>>> from astropy.modeling.models import custom_model
>>> import numpy as np
>>> def sine_model(x, amplitude=1., frequency=1.):
...     return amplitude * np.sin(2 * np.pi * frequency * x)
>>> def sine_deriv(x, amplitude=1., frequency=1.):
...     return 2 * np.pi * amplitude * np.cos(2 * np.pi * frequency * x)
>>> SineModel = custom_model(sine_model, fit_deriv=sine_deriv)


Create an instance of the custom model and evaluate it:

>>> model = SineModel()
>>> model(0.25)
1.0


This model instance can now be used like a usual astropy model.

The next example demonstrates a 2D Moffat function model, and also demonstrates the support for docstrings (this example could also include a derivative, but it has been omitted for simplicity):

>>> @custom_model
... def Moffat2D(x, y, amplitude=1.0, x_0=0.0, y_0=0.0, gamma=1.0,
...            alpha=1.0):
...     """Two dimensional Moffat function."""
...     rr_gg = ((x - x_0) ** 2 + (y - y_0) ** 2) / gamma ** 2
...     return amplitude * (1 + rr_gg) ** (-alpha)
...
>>> print(Moffat2D.__doc__)
Two dimensional Moffat function.
>>> model = Moffat2D()
>>> model(1, 1)
0.3333333333333333