UCP-Networks: A Directed Graphical Representation of Conditional Utilities
UAI '01 Proceedings of the 17th Conference in Uncertainty in Artificial Intelligence
Algorithms for a temporal decoupling problem in multi-agent planning
Eighteenth national conference on Artificial intelligence
Making social choices from individuals' CP-nets
Proceedings of the 6th international joint conference on Autonomous agents and multiagent systems
mCP nets: representing and reasoning with preferences of multiple agents
AAAI'04 Proceedings of the 19th national conference on Artifical intelligence
Extending CP-nets with stronger conditional preference statements
AAAI'04 Proceedings of the 19th national conference on Artifical intelligence
Determining possible and necessary winners under common voting rules given partial orders
AAAI'08 Proceedings of the 23rd national conference on Artificial intelligence - Volume 1
Voting on multiattribute domains with cyclic preferential dependencies
AAAI'08 Proceedings of the 23rd national conference on Artificial intelligence - Volume 1
Journal of Artificial Intelligence Research
Incompleteness and incomparability in preference aggregation
IJCAI'07 Proceedings of the 20th international joint conference on Artifical intelligence
On the foundations of qualitative decision theory
AAAI'96 Proceedings of the thirteenth national conference on Artificial intelligence - Volume 2
Introducing variable importance tradeoffs into CP-nets
UAI'02 Proceedings of the Eighteenth conference on Uncertainty in artificial intelligence
Hi-index | 0.00 |
This work applies the CP-net preference representation to the problem of negotiating optimal joint outcomes, hoping to exploit the CP-net benefit of efficient preferential optimization in multiagent settings. A fundamental challenge in doing so is that acyclic CP-nets only capture an agent's preferences over outcomes qualitatively, as a partial order, making comparisons between agents' strengths of preferences over outcomes problematic. This article presents a plausible (though not the only) strategy to assess outcomes based on their relative positions in the agents' partial orders. Given the ability to compare strength of preference over outcomes, a brute-force search in the space of outcomes can provably yield an optimal (maximin) joint outcome. More importantly, it is shown that the optimal joint outcome can, in principle, be found more efficiently by using a multiagent variation of the CP-net preferential optimization algorithm, provided that the right decisions are made about which agent assigns each variable. Finally, heuristics are developed that find an approximately optimal variable assignment strategy. Empirical evaluation indicates that, relative to the outcome graph search, the new heuristic algorithm based on direct variable assignment achieves exponential speedup, while costing only a small constant factor in solution quality.