About us

History

aGrUM

aGrUM was initially created to support the research activities of the Graphical Models and Decision team at LIP6. What started as a dedicated research tool has gradually evolved into a comprehensive, open-source C++ library designed to serve both the scientific community and industry practitioners working in decision support, probabilistic reasoning, and data science.

The library combines methodological depth with practical efficiency. Researchers benefit from a framework that supports the design, testing, and comparison of advanced models and algorithms, while developers gain access to production-ready tools that can be integrated into real-world applications.

Its key design principles reflect this dual ambition:

• Modern C++20 development: a robust codebase, portable across major compilers (GCC ≥ 8.0, Clang, MSVC), ensuring longevity and cross-platform support
• High performance: carefully optimized data structures and multi-threaded algorithms accelerate both learning and inference, enabling scalability to complex problems and large datasets
• Flexibility and customization: a modular architecture that allows fine-grained control over modeling, learning, and inference strategies—ideal for both experimental research and application-specific tuning
• Balance between theory and practice: algorithms and models are implemented with a focus on scientific soundness, while exposing interfaces that favor usability and integration into larger software systems

To make aGrUM more accessible, various wrappers have been developed, lowering the entry barrier for users who prefer higher-level languages:

pyAgrum

pyAgrum: the primary Python wrapper, offering seamless access to aGrUM’s features from within Python, making it ideal for teaching, prototyping, and data science workflows.

These wrappers are built using SWIG, ensuring consistency with the C++ core while exposing idiomatic interfaces in the target languages.

Mission

By bridging academic research and practical deployment, aGrUM provides a powerful environment for exploring new ideas, validating them through experimentation, and applying them to concrete decision-support and data-driven challenges.

Active Core Contributors

Authors

The following individuals are the authors of this project : design, implement, and maintain the tools that enable users to build, analyze, and manipulate probabilistic or structured graphical models efficiently.

Maintain-and-Deploy Team

The following individuals are the maintainers of this project, responsible for overseeing its development and deployement:

Web Team

The following people maintains, and optimizes the website and its web applications, ensuring functionality, performance, and a seamless user experience, while also promoting the platform and engaging the community through social media outreach.