simpeg.objective_function.ComboObjectiveFunction#

class simpeg.objective_function.ComboObjectiveFunction(objfcts=None, multipliers=None, unpack_on_add=True)[source]#

Bases: BaseObjectiveFunction

Composite for multiple objective functions.

This class allows the creation of an objective function \(\phi\) which is the sum of a list of other objective functions \(\phi_i\). Each objective function has associated with it a multiplier \(c_i\) such that

\[\phi = \sum_{i = 1}^N c_i \phi_i\]

Parameters:

objfctsNone or list of simpeg.objective_function.BaseObjectiveFunction, optional: List containing the objective functions that will live inside the composite class. If None, an empty list will be created.
multipliersNone or list of int, optional: List containing the multipliers for each objective function in objfcts. If None, a list full of ones with the same length as objfcts will be created.
unpack_on_addbool: Whether to unpack the multiple objective functions when adding them to another objective function, or to add them as a whole.

Examples

Build a simple combo objective function:

>>> objective_fun_a = L2ObjectiveFunction(nP=3)
>>> objective_fun_b = L2ObjectiveFunction(nP=3)
>>> combo = ComboObjectiveFunction([objective_fun_a, objective_fun_b], [1, 0.5])
>>> print(len(combo))
2
>>> print(combo.multipliers)
[1, 0.5]

Combo objective functions are also created after adding two objective functions:

>>> combo = 2 * objective_fun_a + 3.5 * objective_fun_b
>>> print(len(combo))
2
>>> print(combo.multipliers)
[2, 3.5]

We could add two combo objective functions as well:

>>> objective_fun_c = L2ObjectiveFunction(nP=3)
>>> objective_fun_d = L2ObjectiveFunction(nP=3)
>>> combo_1 = 4.3 * objective_fun_a + 3 * objective_fun_b
>>> combo_2 = 1.5 * objective_fun_c + 0.5 * objective_fun_d
>>> combo = combo_1 + combo_2
>>> print(len(combo))
4
>>> print(combo.multipliers)
[4.3, 3, 1.5, 0.5]

We can choose to not unpack the objective functions when creating the combo. For example:

>>> objective_fun_a = L2ObjectiveFunction(nP=3)
>>> objective_fun_b = L2ObjectiveFunction(nP=3)
>>> objective_fun_c = L2ObjectiveFunction(nP=3)
>>>
>>> # Create a ComboObjectiveFunction that won't unpack
>>> combo_1 = ComboObjectiveFunction(
...     objfcts=[objective_fun_a, objective_fun_b],
...     multipliers=[0.1, 1.2],
...     unpack_on_add=False,
... )
>>> combo_2 = combo_1 + objective_fun_c
>>> print(len(combo_2))
2

Attributes

`W`	Full weighting matrix for the combo objective function.
`mapping`	Mapping from the model to the quantity evaluated in the object function.
`multipliers`	Multipliers for the objective functions.
`nP`	Number of model parameters.

Methods

`__call__`(m[, f])	Evaluate the objective functions for a given model.
`deriv`(m[, f])	Gradient of the objective function evaluated for the model provided.
`deriv2`(m[, v, f])	Hessian of the objective function evaluated for the model provided.
`get_functions_of_type`(fun_class)	Return objective functions of a given type(s).
`map_class`	alias of `IdentityMap`
`test`([x, num, random_seed])	Run a convergence test on both the first and second derivatives.