SpectralDeltaKernel<T>
Models & Types · AiDotNet.Kernels
Spectral Delta Kernel representing a single spectral component.
Implements: IKernelFunction<T>
For Beginners
The Spectral Delta Kernel is a single component of the Spectral Mixture Kernel. While the SpectralMixtureKernel uses multiple frequency components, this kernel uses just one.
In mathematical terms: k(τ) = σ² × exp(-2π²τ²σ_f²) × cos(2πμτ)
Where:
- τ = |x - x’| is the distance between points
- σ² is the output variance (weight)
- μ is the frequency (how fast the pattern repeats)
- σ_f is the frequency bandwidth (how “spread” the frequency is)
Think of it as a single “note” in the spectrum:
- μ determines the pitch (frequency)
- σ_f determines how pure the tone is (narrow = pure, wide = noisy)
- σ² determines the volume
This is useful when:
- You expect a single dominant periodicity
- You want a simple periodic pattern with decay
- As a building block for more complex spectral kernels
Constructors
| Constructor | Summary |
|---|---|
SpectralDeltaKernel(double frequency, double bandwidth, double variance) |
Initializes a new Spectral Delta Kernel. |
Properties
| Property | Summary |
|---|---|
double Bandwidth { get; } |
Gets the frequency bandwidth. |
double Frequency { get; } |
Gets the mean frequency. |
double Period { get; } |
Gets the period (1/frequency). |
double Variance { get; } |
Gets the output variance. |
Methods
| Method | Summary |
|---|---|
SpectralDeltaKernel<T> FromPeriod(double period, double decayLength, double variance) |
Creates a Spectral Delta Kernel from a known period. |
T Calculate(Vector<T> x1, Vector<T> x2) |
Calculates the spectral delta kernel value. |
double GetPowerSpectralDensity(double f) |
Computes the power spectral density at a given frequency. |