Hit and Run

Small Python package for uniformly sampling convex polytopes with the Hit-and-Run algorithm. Polytopes are given in H-representation (A x <= b).

Includes Polytope for constraints and HitAndRun for generating samples.

Installation

python -m pip install .

Runtime dependencies are declared in pyproject.toml. The core package depends on NumPy; install the optional progress extra to show tqdm progress bars:

python -m pip install ".[progress]"

Usage

import numpy as np

from hitandrun import HitAndRun, Polytope

A = np.array([[1, 0], [-1, 0], [0, 1], [0, -1]], dtype=np.float64)
b = np.array([1, 1, 1, 1], dtype=np.float64)
x0 = np.array([0, 0], dtype=np.float64)

polytope = Polytope(A=A, b=b)
sampler = HitAndRun(polytope=polytope, starting_point=x0)
samples = sampler.get_samples(n_samples=100)

Notebook

See notebooks/hitandrun_usage.ipynb for a plotted walkthrough of defining a polytope, finding a feasible starting point with MinOver, sampling it with HitAndRun, and inspecting empirical marginals. The notebook uses Matplotlib for figures:

python -m pip install ".[progress]" matplotlib

Tests

python -m unittest discover -s tests -v

Contact

Francesc Font-Clos
https://github.com/fontclos
[email protected]