Crate gen_rs

gen_rs is a a library for general-purpose, low-level probabilistic modeling and inference. Modeling and inference are separated by a trait interface, called GenFn.

Any function that implements GenFn (representing a Bayesian model) can use and compose any inference procedure in the standard inference library.

Re-exports

pub use trie::Trie;

pub use address::SplitAddr;

pub use address::normalize_addr;

pub use gfi::Trace;

pub use gfi::GenFn;

pub use gfi::GfDiff;

pub use modeling::dists::u01;

pub use modeling::dists::Distribution;

pub use modeling::dists::bernoulli;

pub use modeling::dists::uniform_continuous;

pub use modeling::dists::uniform;

pub use modeling::dists::uniform_discrete;

pub use modeling::dists::categorical;

pub use modeling::dists::normal;

pub use modeling::dists::mvnormal;

pub use modeling::triefn::TrieFn;

pub use modeling::triefn::TrieFnState;

pub use modeling::triefn::AddrTrie;

pub use modeling::unfold::Unfold;

pub use inference::importance_sampling;

pub use inference::metropolis_hastings;

pub use inference::mh;

pub use inference::mh;

pub use inference::ParticleSystem;

Modules

address

Utilities for parsing addresses (special keys used in the Trie data structure).

gfi

Definition of the Generative Function Interface (GFI).

inference

Standard inference library.

modeling

Distributions and a modeling DSL built on Tries.

trie

Implementations of the Trie data structure, used extensively in modeling::triefn.

Constants

GLOBAL_RNG

Forked PRNG, accessible as a static crate-level thread-local constant. (Use like GLOBAL_RNG.with_borrow_mut(|rng| { ... })).

Functions

logsumexp

For an input vector of [x1, ..., xn], return log(exp(x1) + ... + exp(xn)).