Master Thesis Project
Many Optimization problems can be represented either as a Quadratic�Unconstrained Binary Optimization (QUBO) or as a graph, and selecting the�most appropriate solver for a given problem remains a challenge. The current�methods for solver selection rely on manual expertise and do not fully capture�the complex relationships between the inputs and solver outputs. This�expertise involves having a deep understanding of the problem and the�different solvers available, as well as the ability to match the characteristics of�the problem with the strengths of the solver. However, this expertise can be�limited and may not always lead to the most efficient or effective solver�selection.�The goal of this project is to develop a Recommendation System that uses�Graph Neural Network (GNN) model to accurately predict a good solver, which performs better than the average for an optimization problem,�regardless of its representation as either a QUBO or a graph. The GNN will�take either representation as input and provide a recommendation for the�optimal solver based on its ability to learn from and generalize to previously�seen problems. This proposed solution will improve the efficiency and�effectiveness of solving optimization problems by increasing the accuracy of�solver selection