Distributions Factors
This page lists the built-in methods for creating random variables with various distribution factors. In many cases, you can pass in random variables as arguments e.g. Variable<int> instead of int. For compactness, this is not shown in the syntax below.
These methods provide a convenient short alternative to using Variable<T>.Factor and passing in the factor method, as described on this page. Note also that you can create a random variable with any prior distribution using Variable.Random(IDistribution<T> dist). This is useful if you already have a distribution object, such as the posterior from a previous inference, and wish to create a new random variable with that distribution as its prior.
Discrete Distributions
Distributions where the random variable can take one of a discrete set of states. Where the discrete set of states is indexed by a Range, it should replace a size argument or be passed as an additional parameter.
|
Distribution |
Syntax |
Description |
|
Bernoulli |
Variable.Bernoulli(double probTrue)
|
Creates a boolean random variable from the probability of being true P(true). |
|
Variable.BernoulliFromLogOdds(double logOdds) |
Creates a boolean random variable from the log odds i.e. log P(true)/P(false). Equivalent to Variable.Bernoulli(Variable.Logistic(logOdds)). | |
|
Discrete |
Variable.Discrete(double[] probs) Variable.Discrete(Vector probs) |
Creates an int random variable from supplied array or vector of probabilities (which must add up to 1). The variable can take values between 0 and N-1 where N is the length of the array/vector. |
| Variable.DiscreteFromLogProbs(double[] logProbs) |
Creates an int random variable from supplied array or vector of log probabilities. Equivalent to Variable.Discrete(Variable.Softmax(logProbs)). | |
|
Variable.DiscreteUniform(int size) |
Creates an int random variable which can take values from 0 to size-1 with equal probability. | |
|
Discrete enum
|
Variable.EnumDiscrete<TEnum>(double[] probs) Variable.EnumDiscrete<TEnum>(Vector probs) |
Creates an enum random variable where the probability of each enum value is given by the specified array or vector. TEnum specifies the enum type. |
|
Variable.EnumUniform<TEnum>() |
Creates an enum random variable with equal probability of taking each possible value of the enumeration. TEnum specifies the enum type. | |
|
Binomial |
Variable.Binomial(double probSuccess, int trialCount) |
Creates an int random variable which has a Binomial distribution with the specified probability of success per trial and number of trials. |
| Multinomial |
Variable.Multinomial(Vector probs, int trialCount) |
Creates an int[] random variable array which has a Multinomial distribution with the specified array of probabilities and number of trials. |
| Poisson |
Variable.Poisson(double mean) |
Creates an int random variable which has a Poisson distribution with the specified mean. |
Continuous Distributions
Distributions where the random variable is can take one of a continuous range of values.
|
Distribution |
Syntax |
Description |
| Beta |
Variable.Beta(double trueCount, double falseCount) |
Creates a double random variable with a Beta distribution with the specified counts. |
|
Variable.BetaFromMeanAndVariance(double mean, double variance) |
Creates a double random variable with a Beta distribution of the specified mean and variance. | |
| Gaussian |
Variable.GaussianFromMeanAndPrecision(double mean, double precision) |
Creates a double random variable with a Gaussian distribution of the specified mean and precision (inverse variance). |
|
Variable.GaussianFromMeanAndVariance(double mean, double variance) |
Creates a double random variable with a Gaussian distribution of the specified mean and variance. | |
| Gamma |
Variable.GammaFromShapeAndScale(double shape, double scale) |
Creates a positive double random variable with a Gamma distribution of the specified shape and scale. |
|
Variable.GammaFromShapeAndRate(double shape, double rate) |
Creates a positive double random variable with a Gamma distribution of the specified shape and rate. We now support both stochastic shape and rate, but the support for stochastic shape is experimental. | |
|
Variable.GammaFromMeanAndVariance(double mean, double variance) |
Creates a positive double random variable with a Gamma distribution of the specified mean and variance. |
Multivariate Distributions
Distributions where the random variable is multivariate.
|
Distribution |
Syntax |
Description |
| Dirichlet |
Variable.Dirichlet(double[] u)
Variable.Dirichlet(Vector v) |
Creates a Vector random variable with a Dirichlet distribution with the specified array or vector of pseudo-counts. |
|
Variable.DirichletUniform(int dimension) |
Creates a Vector random variable with a Dirichlet distribution whose pseudo-counts are all set to 1. | |
| Wishart |
Variable.WishartFromShapeAndScale(double shape, PositiveDefiniteMatrix scale) |
Creates a PositiveDefiniteMatrix random variable with a Wishart distribution with the specified shape parameter and scale matrix. |
