AAAI Publications, Workshops at the Twenty-Seventh AAAI Conference on Artificial Intelligence

Font Size: 
Learning CP-Net Preferences Online from User Queries
Joshua T. Guerin, Thomas E. Allen, Judy Goldsmith

Last modified: 2013-06-29


CP-nets offer a compact qualitative representation of human preferences that operate under ceteris paribus ("with all else being equal") semantics. In this paper we present a novel algorithm through which an agent learns the preferences of a user. CP-nets are used to represent such preferences and are learned online through a series of queries generated by the algorithm. Our algorithm builds a CP-net for the user by creating nodes and initializing CPTs, then gradually adding edges and forming more complex CPTs consistent with responses to queries until a confidence parameter is reached.


CP-nets; CP-net learning; preference elicitation

Full Text: PDF