Escaping Cannibalization? Correlation-Robust Pricing for a Unit-Demand Buyer

摘要: We consider a robust version of the revenue maximization problem, where single seller wishes to sell $n$ items unit-demand buyer. In this version, knows buyer's marginal value distribution for each item separately, but not joint distribution, and prices maximize in worst case over all compatible correlation structures. devise computationally efficient (polynomial support size marginals) algorithm that computes worst-case any choice prices. And yet, sharp contrast additive buyer (Carroll, 2017), we show it is NP-hard approximate optimal within factor better than $n^{1/2-\epsilon}$. For special distributions satisfy monotone hazard rate property, how guarantee constant fraction using pricing; pricing equates across possible correlations can be computed efficiently.