flex

The Transformed Density Rejection with Inflection Points (Flex) algorithm can sample from arbitrary distributions given its density function f, its first two derivatives and a partitioning into intervals with at most one inflection point. The partitioning needs to be mutually exclusive and sorted.

auto flex(F0 f0, F1 f1, F2 f2, S c, S[] points, S rho, int maxApproxPoints)
auto flex(Pdf pdf, F0 f0, F1 f1, F2 f2, S c, S[] points, S rho, int maxApproxPoints)
auto flex(F0 f0, F1 f1, F2 f2, S[] cs, S[] points, S rho, int maxApproxPoints)
auto flex(Pdf pdf, F0 f0, F1 f1, F2 f2, S[] cs, S[] points, S rho, int maxApproxPoints)
static if(is(typeof( )))
auto
flex
(
S
Pdf
F0
F1
F2
)
(
in Pdf pdf
,
in F0 f0
,
in F1 f1
,
in F2 f2
,
S[] cs
,
S[] points
,
S rho = 1.1
,
int maxApproxPoints = 1_000
)
if (
isFloatingPoint!S
)
auto flex(Pdf pdf, FlexInterval!S[] intervals)

Parameters

pdf Pdf: probability density function of the distribution
f0 F0: logarithmic pdf
f1 F1: first derivative of logarithmic pdf
f2 F2: second derivative of logarithmic pdf
cs S[]: T_c family array
points S[]: non-overlapping partitioning with at most one inflection point per interval
rho S: efficiency of the Flex algorithm
maxApproxPoints int: maximal number of points to use for the hat/squeeze approximation

Return Value

Type: auto

Flex Generator.

flex

Parameters

Return Value

Meta

Source