Class ModifiedHuberLoss<T>

Namespace

AiDotNet.LossFunctions

Assembly

AiDotNet.dll

Implements the Modified Huber Loss function, a smoother version of the hinge loss.

public class ModifiedHuberLoss<T> : LossFunctionBase<T>, ILossFunction<T>

Type Parameters

T

The numeric type used for calculations (e.g., float, double).

Inheritance

object

LossFunctionBase<T>

ModifiedHuberLoss<T>

Implements

ILossFunction<T>

Inherited Members

LossFunctionBase<T>.NumOps

LossFunctionBase<T>.Engine

LossFunctionBase<T>.ValidateVectorLengths(Vector<T>, Vector<T>)

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Extension Methods

LossFunctionExtensions.ComputeGradient<T>(ILossFunction<T>, Tensor<T>, Tensor<T>)

LossFunctionExtensions.ComputeLoss<T>(ILossFunction<T>, Tensor<T>, Tensor<T>)

Remarks

For Beginners: Modified Huber Loss is a smoother version of the hinge loss that's less sensitive to outliers. It combines quadratic behavior near zero with linear behavior for large negative values.

The formula is:

• For z = -1: max(0, 1 - z)²
• For z < -1: -4 * z

Where z = y * f(x), with y being the true label and f(x) the prediction.

Key properties:

• It's smoother than hinge loss, making optimization easier
• It's more robust to outliers than squared hinge loss
• It combines the benefits of both quadratic and linear losses
• It has a continuous first derivative

Modified Huber Loss is particularly useful for:

• Binary classification problems
• Datasets with noisy labels
• Problems where you want to balance between being sensitive to errors but not overly influenced by extreme mistakes

Methods

CalculateDerivative(Vector<T>, Vector<T>)

Calculates the derivative of the Modified Huber Loss function.

public override Vector<T> CalculateDerivative(Vector<T> predicted, Vector<T> actual)

Parameters

predicted Vector<T>

The predicted values vector.

actual Vector<T>

The actual (ground truth) values vector, typically -1 or 1.

Returns

Vector<T>

A vector containing the derivatives of the modified huber loss with respect to each prediction.

CalculateLoss(Vector<T>, Vector<T>)

Calculates the Modified Huber Loss between predicted and actual values.

public override T CalculateLoss(Vector<T> predicted, Vector<T> actual)

Parameters

predicted Vector<T>

The predicted values vector.

actual Vector<T>

The actual (ground truth) values vector, typically -1 or 1.

Returns

T

The modified huber loss value.

CalculateLossAndGradientGpu(IGpuTensor<T>, IGpuTensor<T>)

Calculates both Modified Huber loss and gradient on GPU in a single efficient pass.

public override (T Loss, IGpuTensor<T> Gradient) CalculateLossAndGradientGpu(IGpuTensor<T> predicted, IGpuTensor<T> actual)

Parameters

predicted IGpuTensor<T>

The predicted GPU tensor from the model.

actual IGpuTensor<T>

The actual (target) GPU tensor.

Returns

(T Loss, IGpuTensor<T> Gradient)

A tuple containing the loss value and gradient tensor.