Regularizers

The regularizers module provides classes for adding regularization terms to objective functions, including constraints, penalties, and priors.

Base Classes

class ionworkspipeline.data_fits.objective_functions.regularizers.Regularizer(fun, regularizer_weight=None)

A regularizer function that takes a dictionary of inputs and returns a value.

Parameters

funpybamm.Symbol or Number

The regularizer expression as a pybamm symbol or a constant number.

regularizer_weightfloat, optional

The weight applied to the regularizer term. Default is 1.0.

evaluate_to_array(inputs: dict) → ndarray

Evaluate the regularizer function with dictionary inputs, returning an array.

Parameters

inputsdict

Dictionary of input values.

Returns

ndarray

Array containing the regularizer value.

evaluate_to_scalar(inputs: dict) → float

Evaluate the regularizer function with dictionary inputs, returning a scalar. The scalar value evaluates the canonical form, which is \((w * f(x))^2\).

Parameters

inputsdict

Dictionary of input values.

Returns

float

Scalar regularizer value.

property fun

callable: The underlying regularizer function.

process_symbol(parameter_values: ParameterValues)

Process the stored symbolic expression into an evaluable function.

Parameters

parameter_valuespybamm.ParameterValues

The parameter values used to resolve parameters in the expression.

property regularizer_weight

float: Weight applied to the regularizer term. Defaults to 1.0.

set_x_to_inputs(x_to_inputs)

Set the function that converts parameter vectors to input dictionaries.

Parameters

x_to_inputscallable

Function that converts a parameter vector to a dictionary of inputs. See ionworkspipeline.DataFit.x_to_inputs() for more information.

to_config() → dict

Convert the regularizer to a configuration dictionary.

property x_to_inputs

callable: Function converting parameter vectors to input dictionaries.

class ionworkspipeline.data_fits.objective_functions.regularizers.RegularizerSet(regularizers: list | tuple | ndarray, x_to_inputs, regularizer_set_weight: float | None = None)

A collection of regularizer terms that can be evaluated together.

Parameters

regularizerslist, tuple or ndarray

Collection of regularizer terms.

x_to_inputscallable

Function converting parameter vectors to input dictionaries.

regularizer_set_weightfloat, optional

Weight applied to the entire regularizer set. Default is 1.0.

Extends: ionworkspipeline.data_fits.objective_functions.regularizers.regularizers.Regularizer

property base_regularizer_type

type: The base type for regularizers in this set.

evaluate_to_array(inputs)

Evaluate all regularizers with dictionary inputs, returning an array.

For consistency with scalar mode (where sum(residuals²) == scalar), the set weight is applied as sqrt(weight) to the residuals. When the optimizer squares the residuals, the effective weight matches the scalar computation.

Parameters

inputsdict

Dictionary of input values.

Returns

ndarray

Array containing all regularizer residuals scaled by sqrt(set_weight).

evaluate_to_scalar(inputs: dict) → float

Evaluate all regularizers in scalar form, returning their weighted sum.

Parameters

inputsdict

Dictionary of input values.

Returns

float

Sum of all regularizer scalar values, multiplied by the set weight.

property regularizer_set_weight

float: Weight applied to the entire regularizer set. Defaults to 1.0.

property regularizers

list: The collection of regularizer terms in this set.

set_x_to_inputs(x_to_inputs)

Set x_to_inputs function for this set and all contained regularizers.

Parameters

x_to_inputscallable

Function converting parameter vectors to input dictionaries.

Submodules