PoissonLikelihood<T>
Models & Types · AiDotNet.GaussianProcesses
Implements the Poisson likelihood for count data.
Implements: ILikelihood<T>
For Beginners
The Poisson likelihood is used for modeling count data (number of events, customer arrivals, defects, etc.).
The latent function f is transformed to a rate using exp: y ~ Poisson(λ), where λ = exp(f)
This means:
- f → +∞ gives high expected counts
- f → -∞ gives low expected counts (near 0)
- f = 0 gives expected count of 1
Properties of Poisson:
- E[y] = λ = exp(f)
- Var[y] = λ = exp(f) (variance equals mean)
Use when:
- Counting discrete events
- Events are independent
- Rate of events can vary smoothly over inputs
Constructors
| Constructor | Summary |
|---|---|
PoissonLikelihood() |
Initializes a new Poisson likelihood. |
Properties
| Property | Summary |
|---|---|
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 count observations. |
T PredictiveVariance(T fMean, T fVariance) |
Computes predictive variance for counts. |
T TransformMean(T f) |
Transforms latent function value to expected count using exp. |
Vector<T> LogLikelihoodGradient(Vector<T> y, Vector<T> f) |
Computes the gradient of the log-likelihood. |
Vector<T> LogLikelihoodHessianDiag(Vector<T> y, Vector<T> f) |
Computes the Hessian diagonal of the log-likelihood. |