PreferenceElicitationinSingleandMultipleUserSettings.ppt

Preference Elicitation in Single and Multiple User Settings;Overview;Preference Elicitation in AI;The Preference Bottleneck;Automated Preference Elicitation;Overview;Constraint-based Decision Problems;Constraint-based Decision Problems;Factored Utilities: GAI Models;Optimization with GAI Models;Difficulties in CBO;Strict Utility Function Uncertainty;f2(L,N) l,n: [2,4] l,n: [1,2];Bayesian Utility Function Uncertainty;f2(L,N) l,n: l,n: …;Query Types;Query Types;Overview;Difficulties with Bound Queries;Local Queries [BB05];Local Standard Gamble Queries;Global Scaling;Bound Query Strategies;Overview;Partial preference informationBayesian uncertainty;Query selection;Bayesian approachMyopic EVOI;Expected value of information;Bayesian approachMyopic EVOI;Local elicitation in GAI models [Braziunas and Boutilier ’05];Experiments;Experiments;Bayesian Elicitation: Future Work;Overview;Minimax Regret: Utility Uncertainty;Why Minimax Regret?*;Why Minimax Regret?;Overview;Computing Max Regret;Minimax Regret in Graphical Models;Constraint Generation;Constraint Generation Performance;Overview;Regret-based Elicitation[Boutilier, Patrascu, Poupart, Schuurmans IJCAI05; AIJ 06];Elicitation Strategies I;Elicitation Strategies II;Elicitation Strategies III;Results (Small Random);Results (Car Rental, Unif);Results (Real Estate, Unif);Results (Large Rand, Unif);Summary of Results;Interpretation;Conclusion – Single User;Questions?;Preference Elicitation in Single and Multiple User Settings Part 2;Overview;Bargaining for a Car;Multiagent PE: Mechanism Design;Basic Social Choice Setup;Basic Mechanism Design;Incentive Compatibility (Truth-telling);Properties;Direct Mechanisms;Groves Schemes;Issues with Classical Mechanism Design;Issues with Classical Mechanism Design;Overview;Partial Revelation Mechanisms;Implications of Roberts;Existing Work on PRMs [Conen,Hudson,Sandholm, Parkes, NisanSegal, BlumrosenNisan];Preference Elicitation in MechDes;Partial Revelati