Rosenbrock.jl

A Julia implementation of the 2D Rosenbrock distribution for testing sampling algorithms.

Based on Hoffman & Ma (2019).

Installation

Make sure that you first have matplotlib installed in your Python environment, as it is required for some plotting functionalities.

pip install matplotlib

Then in Julia:

ENV["PYTHON"] = "/usr/bin/python3"  # Adjust path
using Pkg
Pkg.build("PyCall")
# Restart Julia

] add /path/to/Rosenbrock

Usage

using Rosenbrock
using Distributions

# Create distribution
rb = RosenbrockDistribution(0.0f0, 1.0f0)

# Generate samples
samples = rand(rb, 1000)  # Returns 2×1000 matrix

# Compute log-pdf and gradient
lp = logpdf(rb, samples)  # Extends Distributions.logpdf
grad = gradlogpdf(rb, samples)

println("Sample shape: ", size(samples))  # (2, 1000)
println("Log-pdf shape: ", size(lp))      # (1000,)
println("Gradient shape: ", size(grad))   # (2, 1000)

API

• RosenbrockDistribution(μ, a) - Create distribution with location μ and scale a
• rand(rb, n) - Generate n samples
• logpdf(rb, X) - Compute log probability density
• gradlogpdf(rb, X) - Compute gradient of log probability

Author

Ali Siahkoohi (alisk@ucf.edu), University of Central Florida, 2025

License

MIT