Computational tools for decision-focused learning

The present page contains material for a summer school organized by CERMICS in Fréjus in 2026. The goal is to present two central ideas, which enabled the deep learning revolution and should be used more widely in operations research: automatic differentiation and GPU programming.

These are very rough notes, useful mostly as reminders. If you want to learn this stuff for yourself, read the books by Blondel and Roulet (2025) and Hwu et al. (2022).

Blondel, Mathieu, and Vincent Roulet. 2025. “The Elements of Differentiable Programming.” Pre-published June 24. https://doi.org/10.48550/arXiv.2403.14606.

Hwu, Wen-mei W., David B. Kirk, and Izzat El Hajj. 2022. Programming Massively Parallel Processors: A Hands-on Approach. Morgan Kaufmann. https://books.google.com?id=7H9dEAAAQBAJ.