TanhActivation<T>

Models & Types · AiDotNet.ActivationFunctions

Implements the Hyperbolic Tangent (tanh) activation function for neural networks.

Inheritance: ActivationFunctionBase<T> → TanhActivation<T>

Implements: IFusedActivation, IOutputDerivative<T>

For Beginners

The Hyperbolic Tangent (tanh) activation function is a popular choice in neural networks. It transforms any input value to an output between -1 and 1, creating an S-shaped curve that’s symmetric around the origin.

Key properties of tanh:

  • Outputs values between -1 and 1
  • An input of 0 produces an output of 0
  • Large positive inputs approach +1
  • Large negative inputs approach -1
  • It’s zero-centered, which often helps with learning

When to use tanh:

  • When you need outputs centered around zero
  • For hidden layers in many types of neural networks
  • When dealing with data that naturally has both positive and negative values

One limitation is the “vanishing gradient problem” - for very large or small inputs, the function’s slope becomes very small, which can slow down learning in deep networks.

Example

using AiDotNet.ActivationFunctions;

var activation = new TanhActivation<double>();
double y = activation.Activate(0.5);
Console.WriteLine($"TanhActivation: f(0.5) = {y:F4}");

Constructors

Constructor Summary
TanhActivation()

Properties

Property Summary
bool SupportsGpuTraining { get; } Gets whether Tanh supports GPU-resident training.

Methods

Method Summary
T Activate(T input) Applies the tanh activation function to a tensor of input values.
T Derivative(T input) Calculates the derivative of the tanh function for a tensor input.
Tensor<T> Activate(Tensor<T> input) Applies the tanh activation function to a tensor of input values.
Tensor<T> Derivative(Tensor<T> input) Calculates the derivative of the tanh function for a tensor input.
Tensor<T> DerivativeFromOutput(Tensor<T> output) Computes tanh derivative given the post-activation output: 1 - y².
Vector<T> Activate(Vector<T> input) Applies the tanh activation function to a tensor of input values.
void ForwardGpu(IDirectGpuBackend backend, IGpuBuffer input, IGpuBuffer output, int size) Applies the Tanh activation function on GPU.