Class CompositeRegularizer

Namespace

NeuralNetworks.Regularization

Assembly

NeuralNetworks.dll

Combines multiple regularizers into a single regularization pipeline.

public class CompositeRegularizer : IRegularizer

Inheritance

object

CompositeRegularizer

Implements

IRegularizer

Derived

ElasticNetRegularizer

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

Remarks

Applies regularizers in the order provided for gradient adjustment, penalty calculation, and parameter updates. Penalty reporting is enabled when at least one contained regularizer reports a penalty in the loss, and the reported penalty is the sum of all enabled regularizer penalties.

Constructors

CompositeRegularizer(params BaseRegularizer[])

public CompositeRegularizer(params BaseRegularizer[] regularizers)

Parameters

regularizers BaseRegularizer[]

The regularizers to combine. They are applied in the given order when computing gradients and penalties.

Properties

ReportPenaltyInLoss

Indicates whether penalty is reported as a loss.

public bool ReportPenaltyInLoss { get; }

Property Value

bool

Methods

ApplyGradient<T>(T, T)

Applies a gradient update using the specified parameter values and corresponding gradient values.

public void ApplyGradient<T>(T param, T paramGradient) where T : notnull

Parameters

param T

The parameter value to adjust.

paramGradient T

The gradient of the parameter to adjust.

Type Parameters

T

The type of the parameter and gradient. Must be a non-null type.

Remarks

It calls the ApplyGradientCore(ReadOnlySpan<float>, Span<float>) method of each contained regularizer in the order they were provided, allowing for sequential gradient adjustments.

ReportPenalty<T>(T)

Calculates the total regularization penalty to include in the loss for the provided parameter values.

public float ReportPenalty<T>(T param) where T : notnull

Parameters

param T

The parameter values used to evaluate regularization penalties.

Returns

float

The sum of penalties from regularizers configured to report penalty in loss; otherwise 0.

Type Parameters

T

The parameter container type that can be converted to a read-only span of single-precision values.

Remarks

Returns 0 when penalty reporting in loss is disabled.

ToString()

Returns a string that represents the current object.

public override string ToString()

Returns

string

A string that represents the current object.

UpdateParams<T>(T)

Updates all configured regularizers with the provided parameter values.

public void UpdateParams<T>(T param) where T : notnull

Parameters

param T

The parameter data to convert and pass to each regularizer.

Type Parameters

T

A non-null parameter container type supported by the parameter-to-span conversion routine.

Remarks

Converts the input to a Span<T> and forwards it to each regularizer in sequence.