Closed FriesischScott closed 4 years ago
100% the right move to move from RandomVariableSet
to Copulas
and JointDistributions
.
The to_standard_normal_space
and to_copula_space
are very clean. I didn't know you could do them like that.
What are you using FiniteDifferences.jl for?
FiniteDifferences
is used to compute the gradients needed for LineSampling. Or if you just want to do local sensitivity.
Instead of the
RandomVariableSet
this introduces two new types:JointDistribution
Copula
A
JointDistribution
is built from an array of marginal distributions and a copula of the same dimension. Currently the only implementation ofCopula
is theGaussianCopula
.