Enum Support 

pub enum Support {
    Real,
    Positive,
    Bounded {
        lower: f64,
        upper: f64,
    },
}

The support of a continuous (f64-valued) distribution, as far as an inference kernel needs to know it (FG-N1).

Metropolis-Hastings proposal selection must be a function of the site, not of the value currently at the site: a kernel whose shape depends on the current state is not, in general, invariant for the target. Support is the static description each distribution advertises for that purpose via Distribution::support:

• Support::Real - the whole real line (the default). Proposals are a symmetric Gaussian walk; anything the density rejects scores -inf and is simply rejected, so this is always correct, merely less efficient for a distribution that is really constrained.
• Support::Positive - (0, +inf). Proposals walk in log space with the exact Jacobian correction (FG-02).
• Support::Bounded - a finite interval [lower, upper]. Proposals are a Gaussian walk reflected at both bounds (symmetric, so no Hastings term), which can reach every point of the interval from every other one.

Discrete distributions never consult this and leave the default.

Variants

Real

Unconstrained: (-inf, +inf). Also the conservative default for a distribution that does not describe its support.

Positive

Strictly positive: (0, +inf).

Bounded

A finite interval.

Fields

lower: f64

Lower bound of the interval.

upper: f64

Upper bound of the interval.

Trait Implementations

impl Clone for Support

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Copy for Support

impl Debug for Support

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl PartialEq for Support

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for Support