Daniel Mimouni

I am a PhD student in Applied Mathematics at Mines Paris (CMA) and IFP énergies nouvelles (Applied Mathematics Department), under the supervision of Welington de Oliveira, Paul Malisani and Jiamin Zhu.

Through our objective of solving optimization problems under uncertainty in energy management, we explore, leverage, and enhance techniques in optimal transport, machine learning (mostly RL), convex optimization, and stochastic optimization algorithms.

I am dedicated to combining theoretical research in optimization with practical applications across industries, including energy, biomedical sciences, and technology-driven fields.

Contact: daniel.mimouni@ifpen.fr

🔧 Open-Source Projects

I am committed to reproducible research and open science. All the algorithms developed during my PhD are released as open-source Python packages, each carefully structured, documented, and benchmarked. These repositories reflect both the theoretical depth of my work and its practical implementation in real-world applications such as energy management.

  • Computing-Wasserstein-Barycenters-MAM
    Implementation of the MAM algorithm (published in SIMODS 2024), based on Douglas-Rachford operator splitting. Computes exact Wasserstein barycenters efficiently for both fixed and free supports, in balanced and unbalanced settings.

    plot

  • Constrained-Optimal-Transport
    Extension of the above to constrained barycenter problems (PJOPT 2025). The repo includes methods for convex and non-convex constraints, and illustrates how optimization structure can be preserved in the transport formulation.

    plot

  • Nested_tree_reduction
    Codebase implementing a 10× faster version of the nested Wasserstein-based scenario tree reduction algorithm (submitted to Annals of OR). Offers modular tools to apply it in stochastic optimization workflows.

    plot

  • EMS-RL-DRO
    A full-stack implementation for solving multi-stage energy management problems using Reinforcement Learning and Distributionally Robust Optimization. The project includes scenario generation, benchmarking environments, and is being integrated into IFPEN’s EMS-Lab solver.

Each repository is designed to be plug-and-play, with clean APIs, example notebooks, and thorough documentation. They are actively maintained and open to contributions.

News