Expand description
Full posterior estimation via simulation (Metropolis-Hastings algorithm) and related non-parametric distribution representation (work in progress).
Re-exports§
pub use histogram::*;pub use metropolis::*;
Modules§
- histogram
- Structure to represent one-dimensional empirical distributions non-parametrically (Work in progress).
- metropolis
- Metropolis-Hastings posterior sampler (Work in progress).
Structs§
- EtaTrajectory
- A sequence of natural parameter iterations. The distribution at the current node holds the parameter trajectory of all distributions at the parent nodes, which during optimization or posterior sampling are considered as conditioned or unconditional priors. After optimization/simulation, this trajectory is used to build an approximation to the corresponding posterior entry, which can be retrieved via node.approximate() or node.marginal().
- Sample
- Sample is a collection of 1D posterior marginals, recovered via indexing.
Traits§
- Random
Walk - RandomWalk is implemented by distributions who may maintain a history of changes in the natural parameter scale of its parent(s) node(s).