Class ActivationFunction<TIn, TOut>

Namespace

NeuralNetworks.Operations.ActivationFunctions

Assembly

NeuralNetworks.dll

Defines an abstract 2D activation function operation applied element-wise to two-dimensional data.

public abstract class ActivationFunction<TIn, TOut> : Operation<TIn, TOut> where TIn : notnull where TOut : notnull

Type Parameters

TIn

TOut

Inheritance

object

Operation

Operation<TIn, TOut>

ActivationFunction<TIn, TOut>

Derived

BipolarSigmoid

LeakyReLU2D

LeakyReLU3D

LeakyReLU4D

Linear

ReLU2D

ReLU3D

ReLU4D

Sigmoid

Softplus

Tanh2D

Tanh4D

Inherited Members

Operation<TIn, TOut>.Input

Operation<TIn, TOut>.Output

Operation<TIn, TOut>.Forward(TIn, bool)

Operation<TIn, TOut>.Backward(TOut)

Operation<TIn, TOut>.CalcOutput(bool)

Operation<TIn, TOut>.CalcInputGradient(TOut)

Operation<TIn, TOut>.Forward(object, bool)

Operation<TIn, TOut>.Backward(object)

Operation<TIn, TOut>.GetOutputType()

Operation<TIn, TOut>.GetInputType()

Operation.SetRegistered()

Operation.GetOutputType()

Operation.GetInputType()

Operation.Forward(object, bool)

Operation.Backward(object)

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

Activation functions introduce non-linearity into neural network computations. This base class specifies the contract for operations that transform 2D tensors (e.g., matrices or 2D feature maps), and are typically used in layers processing spatial data without an explicit channel dimension. Concrete implementations should provide the forward activation and, where applicable, support gradient computation for backpropagation by deriving from Operation.