Построение фронта Парето в задачах оптимизации

★ 7.6 · content

cuopt-multi-objective-exploration is a Claude Code skill that traces, completes, and interprets the Pareto frontier across competing objectives by running a sequence of single-objective cuOpt solves using the weighted-sum and ε-constraint methods. It guides the user through building a payoff table across k objectives, normalizing scales by each objective's range, and choosing a scalarization strategy: weighted sum is cheaper but only recovers the convex hull, while ε-constraint sweeps the full frontier including non-convex regions. The skill adds no solver features — it orchestrates LP, MILP, and QP solves already covered by formulation and API skills, keeping the tradeoff decision with the user rather than silently encoding it. Reach for it when a problem has two or more objectives with no agreed-upon weighting, such as cost vs. service level, distance vs. vehicle count, or return vs. risk.