I hold a bachelor’s degree in Mathematics and Computer Science (Double Major) from Université de Montréal. I am currently pursuing a master’s degree in Computer Science at the same institution with specialization in Artificial Intelligence under the supervision of Prof. Andrea Lodi. We are working on improving node selection procedures inside branch-and-bound algorithms using a problem-adaptative, data-driven approach based on imitation learning on a graph neural network agent. Similar research have been done in the past, e.g Gasse & al. on branching, and He & al. on searching, both of which were conclusive. This correlates with Bengio & al., in which according to the authors, machine learning is a promising way of improving combinatorial optimization (CO) and could yield a new generation of combinatorial solvers. These are evidences that suggest this research direction could lead to major advances in CO. Therefore, for my thesis, I will combine machine learning with classical CO to make breakthrough contributions in the new emerging field of data-driven CO.
