Toolkit for Hawkes Processes in Python

Goal

The purpose of Sparklen package is to provide the Python community with a complete suite of cutting-edge tools specifically tailored for the study of exponential Hawkes processes, with a particular focus on high-dimensional framework. It notably features:

• A efficient cluster-based simulation method for generating events.

• A highly versatile and flexible framework for performing inference of multivariate Hawkes process.

• Novel approaches to address the challenge of multiclass classification within the supervised learning framework.

Installation

This section describes how to install the necessary dependencies to set up the package.

1. Install SWIG

Sparklen uses a C++ core code for computationally intensive components, ensuring both efficiency and performance. The binding between C++ and Python is handled through SWIG wrapper code.

So first, you need to install SWIG. Below are the instructions for various platforms.

Anaconda/Miniconda

If you're using Anaconda or Miniconda, install SWIG from the conda-forge channel:

conda install -c conda-forge swig

Linux (Ubuntu/Debian)

On Ubuntu or Debian-based systems, you can install SWIG using apt:

sudo apt update
sudo apt install swig

macOS (Homebrew)

On macOS, you can install SWIG using Homebrew:

brew install swig

Windows

For Windows, follow these steps:

1. Download the latest SWIG release from the SWIG website
2. Add the SWIG folder to your system's PATH environment variable

If you are using Chocolatey you can also install SWIG by running:

choco install swig

2. Get the Source Code

Clone the repository to get the latest version of the source code:

git clone https://github.com/romain-e-lacoste/sparklen.git
cd sparklen

3. Install the Package

It's recommended to set up a dedicated Python environment (e.g., using venv or conda). Once your environment is ready, install the package by running:

pip install .

Citing this work

If you found this package useful, please consider citing it in your work:

@article{lacoste2025sparkle,
      title={Sparkle: A Statistical Learning Toolkit for High-Dimensional Hawkes Processes in Python}, 
      author={Lacoste, Romain E.},
      year={2025},
      eprint={2502.18979},
      archivePrefix={arXiv},
      primaryClass={stat.ME},
      url={https://arxiv.org/abs/2502.18979}, 
}

Acknowledgement

This work has been supported by the Chaire “Modélisation Mathématique et Biodiversité” of Veolia-École polytechnique-Museum national d’Histoire naturelle-Fondation X