Deutsch-französisches Graduiertenkolleg: Statistisches Lernen für komplexe stochastische Prozesse/ Apprentissage statistique pour des processus stochastiques complexes

This Franco-German doctoral college project brings together two French partner institutions from Occitanie (Toulouse and Montpellier) and two German partner institutions from Berlin-Brandenburg (Potsdam and Berlin). The project focuses on research questions at the intersection of probability, statistics, and machine learning. It addresses complex stochastic processes in various fields such as ecological modeling, physics, and optimization. These processes include branching processes, allele frequency spectrum modeling, random energy models, exclusion processes, percolation, and bandits. The project also involves analyzing long-term behaviors, deviations, and mixing times of Markovian and non-Markovian processes. Furthermore, it integrates these complex stochastic processes into algorithms for unsupervised learning tasks like clustering, ranking, and seriation. Additionally, the project explores sequential learning on these processes, such as multi-armed bandit models and reinforcement learning.
The project involves participants from the four partner institutions who cover different aspects of the project in a complementary manner. Its main objective is to foster scientific education through the exchange of doctoral candidates. The project leaders represent the thematic diversity of the project and offer a probabilist and a statistician as the primary point of contact for doctoral candidates in each country. This setup provides doctoral candidates with valuable mobility opportunities in the project's two main axes. The doctoral candidates and postdoctoral participants will broaden their scientific horizon, by offering them joint supervision by scientific teams from distinct countries.