Special Distributions
Differ from the behavior of Distributions.jl.
KernelDistributions.Circular
KernelDistributions.CircularUniform
KernelDistributions.SmoothExponential
KernelDistributions.TailUniform
Bijectors.logabsdetjac
Bijectors.transform
Bijectors.transform
KernelDistributions.Circular
— TypeCircular
Transform ℝ → [0,2π)
KernelDistributions.CircularUniform
— TypeCircularUniform
Generates uniform random numbers ∈ [0,2π]. If the lower/upper bound is exceeded the value continues at the other bound (2π+0.1=0.1).
Bijectors.logabsdetjac
— Methodlogabsdetjac(Circular, x)
mod2pi will not be zero for n2π, thus the discontinuity will not be reached. Thus, the log Jacobian is always 0.
Bijectors.transform
— Methodtransform(Circular, y)
Uses mod2pi
to transform ℝ to [0,2π].
Bijectors.transform
— Methodtransform(Circular, x)
Transform from [0,2π] to ℝ. In theory inverse of mod does not exist, in practice the same value is returned, since [0,2π] ∈ ℝ
KernelDistributions.SmoothExponential
— TypeSmoothExponential(min, max, β, σ)
Smooth truncated exponential distribution by convolving the exponential with a normal distribution: Smooth = Exp ⋆ Normal This results in smooth min and max limits and a definition on ℝ instead of ℝ⁺
Does not support truncated
of Distributions.jl since it is a smooth truncation of the exponential distribution.
KernelDistributions.TailUniform
— TypeTailUniform(min, max)
Acts like a uniform distribution of [min,max] but ignores outliers and always returns 1/(max-min) as probability.