Amine Bennouna

Massachusetts Institute of Technology
Laboratory for Information & Decision Systems

Contact: amineben [at] mit.edu
CV, Google Scholar, LinkedIn


Hello! I am a postdoctoral researcher at the MIT Laboratory for Information & Decision Systems (LIDS). I will join the Kellogg School of Management at Northwestern University as an Assistant Professor in Operations starting July 2025.

I completed my PhD at the MIT Operations Research Center in 2024, advised by Prof. Bart Van Parys. Prior to joining MIT, I graduated from Ecole Polytechnique in 2019 majoring in Applied Mathematics.


headshot

My research focuses on understanding how machines, or artificial intelligence (AI), learn to make decisions. Specifically, I work on developing novel learning algorithms to enable efficient, data-driven decision-making while emphasizing key reliability attributes. Enhancing these learning algorithms has significant practical implications, driven by the rapid adoption of AI, and is based on fascinating mathematical models. My research involves theoretical and algorithmic advancements, leveraging tools from probability theory, optimization, and geometry.
Keywords include: Stochastic Optimization, Distributionally Robust Optimization, Machine Learning, Reinforcement Learning, Data-driven Decision-making, Experimental Design.

Recent News
  • New paper: From Distributional Robustness to Robust Statistics: A Confidence Sets Perspective.

    We establish a natural connection between distributionally robust optimization and classical robust statistics, the two major frameworks for robustness that often seem to follow fundamentally different paradigms.

  • Published paper: Learning the Minimal Representation of a Continuous State-Space Markov Decision Process from Transition Data is published in Management Science.

    We study here how to learn the simplest MDP model that explains sequential data observed from a dynamic system. This is a very intriguing statistical learning problem theoretically, with important practical implications for interpretability in ML.

Papers
In preparation:

  • Robust Two-Stage Optimization with Covariate Data
    with Bart Van Parys & Julien Pinede.
Teaching
  • Optimization Methods, MIT 15.093/6.255 | Head Teaching Assistant
    Graduate (Masters, PhDs, MBAn, MBA), Fall 2021. (180 students)
  • Optimization Methods, MIT 15.093/6.255 | Teaching Assistant
    Graduate (Masters, PhDs, MBAn, MBA), Fall 2020. (120 students)
  • The Analytics Edge, MIT 15.071 | Guest Lecturer
    Graduate (MBA), Fall 2023. (80 students)
  • The Advanced Analytics Edge, MIT 15.072 | Guest Lecturer
    Graduate (MBAn), Fall 2023. (100 students)
  • Classes Préparatoires Instructor (Louis-le-Grand, Saint Louis, Henri IV)
    Instructor and examiner in advanced mathematics for undergraduate students of top French classes préparatoires
Talks
The background is a typical geometric motif of Moroccan's mosaic tilework. These beautiful geometric patterns, found everywhere in Morocco, date back to the 10th century. Beyond their beauty, they are facinating mathematical objects. (Credit: Zellij gallery)