Gbètondji Dovonon

Graphs are natural representations for data in a wide array of applications, ranging from materials science to algorithm design. Often however, these graphs are not available in data: e.g., a potential library of ideal catalysts may not have known synthesis graphs required to construct them. In this project our goal is to devise models and optimization algorithms to learn such graphs implicitly for maximizing arbitrary objectives (e.g., reaction yield). This will involve theoretical developments in graph modelling, discrete optimization, and topological analysis. As the project is aimed at fundamental challenges in graph learning, the resulting methods will be useful across the set of applications mentioned above.