Class MaxNormRegularizer

Namespace

NeuralNetworks.Regularization

Assembly

NeuralNetworks.dll

Applies max-norm regularization by scaling a parameter vector when its L2 norm exceeds a specified limit.

public class MaxNormRegularizer : BaseRegularizer, IRegularizer

Inheritance

object

BaseRegularizer

MaxNormRegularizer

Implements

IRegularizer

Inherited Members

BaseRegularizer.ReportPenaltyInLoss

BaseRegularizer.ApplyGradient<T>(T, T)

BaseRegularizer.ReportPenalty<T>(T)

BaseRegularizer.UpdateParams<T>(T)

BaseRegularizer.ApplyGradientCore(ReadOnlySpan<float>, Span<float>)

BaseRegularizer.CalculatePenaltyCore(ReadOnlySpan<float>)

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

Remarks

If the vector norm is greater than the configured limit, all elements are scaled by the same factor so the resulting norm equals the limit.

Constructors

MaxNormRegularizer(float)

public MaxNormRegularizer(float maxNorm)

Parameters

maxNorm float

The maximum allowed L2 norm for the parameter vector.

Methods

ToString()

Returns a string that represents the current object.

public override string ToString()

Returns

string

A string that represents the current object.

UpdateParamsCore(Span<float>)

Scales the parameter vector in place when its Euclidean norm exceeds the configured maximum norm.

protected override void UpdateParamsCore(Span<float> paramSpan)

Parameters

paramSpan Span<float>

A span containing the parameter values to evaluate and, if needed, scale.

Remarks

Computes the L2 norm of the span and applies a uniform scaling factor so the resulting norm equals the maximum norm while preserving the vector direction.