GaussianActivation<T>

Models & Types · AiDotNet.ActivationFunctions

Implements the Gaussian activation function for neural networks.

Inheritance: ActivationFunctionBase<T> → GaussianActivation<T>

Implements: IFusedActivation

For Beginners

The Gaussian activation function is based on the bell-shaped curve that you might recognize from statistics (the “normal distribution” or “bell curve”).

Key properties of the Gaussian activation function:

  • It outputs values between 0 and 1
  • The highest output (1) occurs when the input is 0
  • As inputs move away from 0 (either positive or negative), the output approaches 0
  • It’s symmetric around the y-axis (f(-x) = f(x))

Unlike many other activation functions, Gaussian responds strongly to inputs near zero and weakly to inputs far from zero in either direction. This makes it useful for:

  • Radial Basis Function (RBF) networks
  • Pattern recognition tasks
  • Problems where distance from a central point is important

The mathematical formula is: f(x) = exp(-x²)

Example

using AiDotNet.ActivationFunctions;

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

Constructors

Constructor Summary
GaussianActivation()

Methods

Method Summary
T Activate(T input) Applies Gaussian to a tensor via engine primitives so the gradient tape records every step of exp(-x²).
T Derivative(T input) Calculates the derivative of the Gaussian function for a single input value.
Tensor<T> Activate(Tensor<T> input) Applies Gaussian to a tensor via engine primitives so the gradient tape records every step of exp(-x²).