cristal.helper_classes.regularizer module

Regularization methods for score functions.

class cristal.helper_classes.regularizer.CombRegularizer

Bases: BaseRegularizer

Regularizer for the combinatorial regularization method : \(reg = \begin{pmatrix} d + n \\ d \end{pmatrix}\).

static regularizer(n, d, C)

Compute the combinatorial regularization factor.

Return type:

float

class cristal.helper_classes.regularizer.ConstantRegularizer

Bases: BaseRegularizer

Regularizer that applies a constant factor C : : \(reg = C\).

static regularizer(n, d, C)

Compute the constant regularization factor.

Return type:

float

class cristal.helper_classes.regularizer.IMPLEMENTED_REGULARIZATION_OPTIONS(*values)

Bases: Enum

The implemented regularization classes.

COMB = <class 'cristal.helper_classes.regularizer.CombRegularizer'>

CONSTANT = <class 'cristal.helper_classes.regularizer.ConstantRegularizer'>

VU = <class 'cristal.helper_classes.regularizer.VuRegularizer'>

VU_C = <class 'cristal.helper_classes.regularizer.VuCRegularizer'>

class cristal.helper_classes.regularizer.VuCRegularizer

Bases: BaseRegularizer

Regularizer for the Vu_C regularization method : \(reg = \frac{n^{3d/2}}{C}\).

See https://arxiv.org/abs/1910.14458 for more details.

static regularizer(n, d, C)

Compute the Vu_C regularization factor.

Return type:

float

class cristal.helper_classes.regularizer.VuRegularizer

Bases: BaseRegularizer

Regularizer for the Vu regularization method : \(reg = n^{3d/2}\).

See https://arxiv.org/abs/1910.14458 for more details.

static regularizer(n, d, C)

Compute the Vu regularization factor.

Return type:

float