Answers are grounded in this paper and cite their sources.
A central question in hybrid quantum–classical optimization is how to assign the classical and quantum components according to the structure of the problem, rather than attempting to transfer the full optimization model to a quantum device. We investigate this question on static and finite-horizon dynamic restricted vertex minimum multicut, an NP-hard network-isolation problem with exponentially many path constraints. A classical row-generation layer handles graph separation, feasibility verification, and constraint generation, while quantum or quantum-inspired samplers are used only for the current restricted optimization problem. We develop two such algorithms: an exact-slack QUBO formulation sampled by simulated annealing, and a fixed-register CVaR-VQE formulation in which generated constraints modify a classical loss without enlarging the static quantum register. Exact MILP row generation provides the certifying reference. Experiments on SNDlib-derived telecom instances show that both approaches can be integrated consistently within this workflow, while classical-sampler controls do not establish a quantum-specific performance advantage. The dynamic study further shows how temporal coupling creates a distinct resource bottleneck, motivating decomposed formulations for larger horizons.
No takes yet. Share an insight, caveat, or question.
Abbassi et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: