GaussianLikelihood<T>

Models & Types · AiDotNet.GaussianProcesses

Implements the Gaussian (Normal) likelihood for regression.

Implements: ILikelihood<T>

For Beginners

The Gaussian likelihood is the standard choice for regression. It assumes observations are the true function value plus Gaussian noise:

y = f(x) + ε, where ε ~ N(0, σ²)

This means:

  • Errors are normally distributed
  • Errors have constant variance (homoscedastic)
  • Errors are independent

The noise variance σ² is a hyperparameter that:

  • Large σ² → Smoother fit, more uncertainty
  • Small σ² → Interpolates data more closely

Gaussian likelihood allows exact GP inference (no approximations needed).

Constructors

Constructor Summary
GaussianLikelihood(double noiseVariance) Initializes a new Gaussian likelihood.

Properties

Property Summary
T NoiseVariance { get; } Gets the noise variance.
string Name { get; } Gets the name of this likelihood.

Methods

Method Summary
T LogLikelihood(Vector<T> y, Vector<T> f) Computes the log-likelihood of observations given latent function values.
T PredictiveVariance(T fMean, T fVariance) Computes predictive variance (adds noise variance to latent variance).
T TransformMean(T f) For Gaussian likelihood, the mean is just f (identity transform).
Vector<T> LogLikelihoodGradient(Vector<T> y, Vector<T> f) Computes the gradient of the log-likelihood with respect to f.
Vector<T> LogLikelihoodHessianDiag(Vector<T> y, Vector<T> f) Computes the Hessian diagonal of the log-likelihood.