Marie-Anne Poursat: educating the mathematicians of the future

"I always wanted to teach," reveals Marie-Anne Poursat. This determination led her to ENS Paris-Saclay, where after a 2-year preparatory scientific course for a competitive exam, she took the 'agrégation' teaching qualification in mathematics. Curious about mathematical applications, she decided to continue her studies with a master's degree in statistics and stochastic models at Université Paris-Saclay and completed an internship at the biometrics laboratory at the National Institute for Agricultural Research (INRA), which is now the National Research Institute for Agriculture, Food and Environment (INRAE). Excited by this stimulating environment, she embarked on a career in research. "This wasn't just statistical research specific to the discipline, but also involved numerous collaborations with the INRAE teams. I was spending time with passionate people from all walks of life. I liked it so much I stayed." Enjoying the status of research assistant at INRAE, she completed a thesis and in 1992, joined INRAE as a research fellow.
 

Overcoming conceptual bottlenecks

Marie-Anne Poursat's early research immersed her in mathematical statistics, during the formative period of non-parametric methods. "I worked on statistical calibration applied to biological assays. Instead of using a rigid statistical model, we aimed to let the data speak for itself. For biological assays, this meant finding the concentration of a molecule or the efficacy of a treatment in the presence of an unknown dose-response curve, which therefore had to be estimated." Ultimately, the objective was to make the model more robust.

Publishing in the prestigious Annals of Statistics journal, the young mathematician was rapidly invited to several international conferences. "I remember a conference in Toronto where the quality of my presentation surprised the other guest speakers, who had mistaken me for the receptionist," she laughs. In 1994, she completed a post-doctorate at the National Institute of Statistics and Economic Studies (INSEE) research centre. During this period, Marie-Anne Poursat was also involved in educating her INRA colleagues about statistical methods, and was increasingly working with teams in various fields, providing them with her specific expertise. "Whether they were agronomists or biologists, the idea was to work on mathematical models applied to their subject. For me, this wasn't just calculation, but also formalising a question: nothing motivates me more than solving conceptual bottlenecks!"
 

From INRA to University

In 2000, Marie-Anne Poursat's university colleagues asked her to develop a professional master's programme in mathematical engineering, leading to a career change from research fellow at INRA to lecturer at Université Paris-Saclay. "I had a free hand to create data mining courses and develop the corporate network." Since 2001, as part of the Orsay Mathematics Laboratory (LMO - Univ. Paris-Saclay/CNRS), she has been working on new statistical methods for phylogenetic reconstruction, in partnership with fellow bioinformaticians. "Phylogenetics reconstructs the evolution of living organisms through the evolution of their DNA sequences, along a ‘phylogenetic’ tree. At the time, models were evolving at constant speeds. The aim was to introduce heterogeneous evolutionary speeds." From 2003 to 2006, the lecturer coordinated a project funded by the French government's “Action Concertée Incitative” (ACI) programme, "New mathematics interfaces", intended to introduce non-parametric methods for selecting or comparing phylogenetic trees from different models.
 

Research and teaching, a daily challenge

Between 2007 and 2014, Marie-Anne Poursat was in charge of the second year of the master's programme in mathematical engineering. At the same time, she continued to collaborate with bioinformatics colleagues. Her research into new machine learning methods for large-scale data led to several publications. "This was predominantly expert work, but for me, it provided fundamental methodological insights." The mathematician also returned to her favoured early themes through a collaboration with Inria, firstly in 2011 with the POPIX team, based at the LMO, "on variable selection methods in population pharmacology models", then from 2018 with the CELESTE project-team "on the statistical study of rankings of big biological data".
 

The creation of double undergraduate degrees

In 2015, Marie-Anne Poursat took responsibility for the third year of the mathematics undergraduate degree. She subsequently discovered that her students did not have access to the master's degree in mathematics, which was reserved for students from the magister postgraduate degree and 2-year preparatory courses. Together with her colleague Nicolas Burq, she approached the University's mathematics department and proposed creating attractive undergraduate degrees. These would enable motivated secondary school students to be educated under the best possible conditions for master's degrees, particularly in mathematics. This resulted in the double undergraduate degrees (computer science-mathematics, mathematics-physics, mathematics-life sciences, economics-mathematics) which would later become Université Paris-Saclay's double degrees. From 2017 to 2018, Marie-Anne Poursat managed the development of programme models and coordinated the new courses. "That's what I like about the University: you can be creative, as long as you can prove the relevance of the project."
 

Artificial intelligence: the future of mathematics

Marie-Anne Poursat, who now co-manages the double undergraduate degree in computer science and mathematics, found once again that her students were unable to pursue a master's degree in mathematics at Université Paris-Saclay, due to a lack of places. She convinced her colleagues to create the Mathematics and Artificial Intelligence study path, open to double-degree students. She is now in charge of the first year of the master's programme. "Starting with twenty students in September 2022, it has been a great success, with promising career opportunities for graduates," she enthuses. However, she is also observing rapid change in the sector. "To secure an interesting position in the field of AI, a PhD is increasingly required. This is why many of these students continued last year with an Industrial Research Training Agreement (CIFRE) or academic thesis."

Another issue is the low number of girls who enrol. "This year, there are only three girls out of 23 students," she says sadly. Marie-Anne Poursat is advocating for greater gender diversity in scientific courses, stressing that "AI professions are very cross-disciplinary and we need to diversify our approaches". She also points to the lack of female role models in this sector and selective recruitment, "a factor leading to self-censorship by girls". Over the next few years, she wants to help correct this bias. "Maths and AI are for girls too!" she concludes.