Improved prophet inequalities for combinatorial welfare maximization with (approximately) subadditive agents
作者:
Highlights:
•
摘要
We give a framework for designing prophet inequalities for combinatorial welfare maximization. Instantiated with different parameters, our framework implies (1) an O(logm/loglogm)-competitive prophet inequality for subadditive agents, (2) an O(Dlogm/loglogm)-competitive prophet inequality for D-approximately subadditive agents, where D∈{1,…,m−1} measures the maximum number of items that complement each other, and (3) as a byproduct, an O(1)-competitive prophet inequality for submodular or fractionally subadditive (a.k.a. XOS) agents, matching the optimal ratio asymptotically. Our framework is computationally efficient given sample access to the prior, value queries and demand queries.
论文关键词:Prophet inequalities,Combinatorial welfare maximization,(Approximate) subadditivity
论文评审过程:Received 2 January 2021, Revised 29 July 2021, Accepted 18 August 2021, Available online 1 September 2021, Version of Record 8 September 2021.
论文官网地址:https://doi.org/10.1016/j.jcss.2021.08.003