Class NeuralNetworkDerivatives<T>

Namespace

AiDotNet.Autodiff

Assembly

AiDotNet.dll

Provides first- and second-order derivatives for neural networks with safe fallbacks.

public static class NeuralNetworkDerivatives<T>

Type Parameters

T

The numeric type used for calculations.

Inheritance

object

NeuralNetworkDerivatives<T>

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Methods

ComputeDerivatives(NeuralNetworkBase<T>, T[], int)

Computes first and second derivatives for a feedforward network at a single input point. Falls back to finite differences when analytic derivatives are unavailable.

public static PDEDerivatives<T> ComputeDerivatives(NeuralNetworkBase<T> network, T[] inputs, int outputDim)

Parameters

network NeuralNetworkBase<T>

inputs T[]

outputDim int

Returns

PDEDerivatives<T>

ComputeGradient(NeuralNetworkBase<T>, T[], int)

Computes first derivatives (Jacobian) for a network output with autodiff-first fallback.

public static T[,] ComputeGradient(NeuralNetworkBase<T> network, T[] inputs, int outputDim)

Parameters

network NeuralNetworkBase<T>

inputs T[]

outputDim int

Returns

T[,]

ComputeHessian(NeuralNetworkBase<T>, T[], int)

Computes the Hessian for a scalar output index.

public static T[,] ComputeHessian(NeuralNetworkBase<T> network, T[] inputs, int outputIndex = 0)

Parameters

network NeuralNetworkBase<T>

inputs T[]

outputIndex int

Returns

T[,]