Optimizing QUBO generation parameters for NP problems and their impact on D-Wave convergence

NP problems are closely related to practical optimization challenges but often suffer from exponential increases in computation time as problem sizes grow. Quantum annealing offers a promising approach to solve NP problems faster than classical methods, requiring cost functions and constraints to be expressed in QUBO format. However, setting QUBO coefficients often relies on empirical methods, particularly in scheduling problems where large values and intuition are needed. In this study, we analyzed QUBO formulas for three coloring-related problems: the graph coloring problem, the clique vertex cover problem, and the integer-length job scheduling problem. We identified the need for independent parameters for complex problems and derived relationships between formula characteristics and optimal QUBO parameter values. Using D-Wave quantum annealing, we validated these parameters and visualized the effects of parameter changes on states and eigenvalues in small spin problems. Additionally, we demonstrated how independent Ising coefficients enhance convergence to correct states based on optimal parameter adjustments.