I believe in open science principles to share software efficiently
A Python package for maximum likelihood estimation of parametric models, with a special focus on discrete choice models.
Professor Michel Bierlaire believes in open science principles to share software efficiently. This software has always been developed to be shared (initially with Prof. Bierlaire’s PhD students), and therefore rigorously packaged and documented. Since its first release in the late 90s, Biogeme has been available, maintained and regularly updated to match the latest technologies. For example, turning Biogeme into a Python package in 2018 was a strategic choice to integrate it into a language that is known as the workhorse of data analysis, and enjoys a large community of users and contributors.
Biogeme is used for teaching discrete choice at several world-renowned universities, including MIT, TU Munich, NTU Athens, and the Technion. The conference paper introducing the software has been cited more than 1300 times, making it the most cited paper by Professor Bierlaire. The active community of users exchanges via a Google group of over 600 members.
Professor Bierlaire manages alone the core development, the maintenance and the user support, because he considers that the involvement of PhD students in this project has little academic added value for them. The two tasks that take the most time are
Packaging for distribution with each new release: extensive testing and documentation, and ensuring operating system compatibility (especially on Windows, which most of the community uses but where “it is a nightmare to deploy”).
User support, although most of the questions are simple, is part of Michel Bierlaire’s daily email activity, and some of the tricky ones help identify bugs and improve the software.