IParametricDistribution<T>

Interfaces · AiDotNet.Distributions

Defines a parametric probability distribution with learnable parameters.

Known implementations: ISamplingDistribution<T>

For Beginners

A parametric distribution is like a template for describing uncertainty. For example, a Normal distribution is defined by just two numbers: mean (center) and variance (spread). Once you know these parameters, you know everything about the distribution.

How It Works

Parametric distributions are fully specified by a fixed set of parameters (e.g., mean and variance for Normal, shape and rate for Gamma). This interface provides methods for computing probability densities, cumulative distributions, and gradients needed for gradient-based learning.

Properties

Property Summary
T Mean { get; } Gets the mean (expected value) of the distribution.
T StdDev { get; } Gets the standard deviation of the distribution.
T Variance { get; } Gets the variance of the distribution.
Vector<T> Parameters { get; set; } Gets or sets the distribution parameters as a vector.
int NumParameters { get; } Gets the number of parameters that define this distribution.
string[] ParameterNames { get; } Gets the parameter names for this distribution.

Methods

Method Summary
IParametricDistribution<T> Clone() Creates a copy of this distribution with the same parameters.
Matrix<T> FisherInformation() Computes the Fisher Information Matrix for the distribution.
T Cdf(T x) Computes the cumulative distribution function (CDF) at point x.
T InverseCdf(T p) Computes the inverse CDF (quantile function) for probability p.
T LogPdf(T x) Computes the log probability density function at point x.
T Pdf(T x) Computes the probability density function (PDF) at point x.
Vector<T> GradLogPdf(T x) Computes the gradient of the log PDF with respect to each parameter.