SQRBFActivation<T>
Models & Types · AiDotNet.ActivationFunctions
Implements the Squared Radial Basis Function (SQRBF) activation function.
Inheritance: ActivationFunctionBase<T> → SQRBFActivation<T>
Implements: IFusedActivation
For Beginners
The Squared Radial Basis Function (SQRBF) is an activation function that produces a bell-shaped curve. Unlike functions like ReLU or Sigmoid that are used in standard neural networks, SQRBF is commonly used in Radial Basis Function Networks (RBFNs).
Think of SQRBF like a “proximity detector” - it gives its highest output (1.0) when the input is exactly 0, and progressively smaller outputs as the input moves away from 0 in either direction (positive or negative). The ß parameter controls how quickly the output drops off as you move away from 0:
- A larger ß makes the bell curve narrower (drops off quickly)
- A smaller ß makes the bell curve wider (drops off slowly)
This is useful in machine learning when you want to measure how close an input is to a specific reference point.
How It Works
The SQRBF activation function is defined as f(x) = exp(-ß * x²), where ß is a parameter that controls the width of the Gaussian bell curve. This function outputs values between 0 and 1, with the maximum value of 1 occurring when the input is 0, and values approaching 0 as the input moves away from 0 in either direction.
Example
using AiDotNet.ActivationFunctions;
var activation = new SQRBFActivation<double>();
double y = activation.Activate(0.5);
Console.WriteLine($"SQRBFActivation: f(0.5) = {y:F4}");
Constructors
| Constructor | Summary |
|---|---|
SQRBFActivation(double beta) |
Initializes a new instance of the SQRBFActivation class with the specified beta parameter. |
Methods
| Method | Summary |
|---|---|
T Activate(T input) |
Applies SQRBF to a tensor via engine primitives so the gradient tape records every step. |
T Derivative(T input) |
Calculates the derivative of the SQRBF function for a given input value. |
Tensor<T> Activate(Tensor<T> input) |
Applies SQRBF to a tensor via engine primitives so the gradient tape records every step. |