hybrid_rosenbrock

A python implementation of https://arxiv.org/abs/1903.09556

Motivation

It is important to have reliable benchmarks to check against in order to compare Bayesian inference algorithms and quantify their performance. The hybrid Rosenbrock distribution is a benchmark that

1. may be sampled directly,
2. may be extended to an arbitrary dimension,
3. may be tuned to increased or decreased difficulty easily,
4. resembles many practical problems of interest (e.g, https://arxiv.org/abs/2106.15163).

For these reasons the hybrid Rosenbrock is a very desirable benchmark for MCMC testing.

Getting started