Simulation-Based Inference: A Practical Guide

This repository contains code to regenerate the figures and numerical results of the practical guide on simulation-based inference . Each figure directory under paper/ contains the simulator configuration, inference scripts / notebooks, and post-processing needed to produce the corresponding figure panel(s).

Repository Structure

• paper/fig*/ – Figure-specific assets (notebooks, helper modules, generated fig/ and svg/ outputs).
• tasks.py – Invoke tasks for converting exported SVG figures to PDF/PNG via Inkscape.

Installation

We recommend using uv for installing the dependencies. After installing uv, run

uv venv -p python3.11
source .venv/bin/activate
uv sync

Reproducing Figures

1. Open the target paper/figX_*/ notebook (e.g. 01_create_figure.ipynb).
2. Execute cells to (a) run / load simulations, (b) run SBI methods, (c) assemble plots into fig/.

Large Files / Pre‑simulated Data (Git LFS)

Some figures rely on pre‑simulated artifacts tracked with Git LFS (e.g. DDM, gravitational wave); after cloning run git lfs install && git lfs pull to fetch binaries (otherwise you only have pointer text files). Only add new large files if essential for reproducibility and track them first with git lfs track before committing.

Contributing

Small improvements (tests, reproducibility, docs) are welcome via pull request.