pymc-bart-rs 0.1.0

No project description provided

bart-rs

Rust implementation of PyMC-BART.

Usage

...

Modifications

The core Particle Gibbs (PG) sampling algorithm for Bayesian Additive Regression Trees (BART) remains the same in this Rust implementation. What differs is the choice of data structure to represent the Binary Decision Tree.

A DecisionTree structure is implemented as a number of parallel vectors. The i-th element of each vector holds information about node i. Node 0 is the tree's root. Some of the arrays only apply to either leaves or split nodes. In this case, the values of the nodes of the other vector is arbitrary. For example, feature and threshold vectors only apply to split nodes. The values for leaf nodes in these arrays are therefore arbitrary.

Design

In this section, the architecture of bart-rs is given.

TODO...

Seeding RNGs

The implementation of BART utilizes randomness in the growing of trees. The thread_rng function from the randr_distr crate provides a thread-local random number generator that is automatically seeded by the operating system or environment, ensuring that it is unique for each thread and run of the program. Therefore, we do not explicitly set a specific seed, and expect different values, e.g. sampled values from a Normal distribution, each time the program is ran.