TY - GEN

T1 - Directional decomposition of multiattribute utility functions

AU - Brafman, Ronen I.

AU - Engel, Yagil

PY - 2009/12/14

Y1 - 2009/12/14

N2 - Several schemes have been proposed for compactly representing multiattribute utility functions, yet none seems to achieve the level of success achieved by Bayesian and Markov models for probability distributions. In an attempt to bridge the gap, we propose a new representation for utility functions which follows its probabilistic analog to a greater extent. Starting from a simple definition of marginal utility by utilizing reference values, we define a notion of conditional utility which satisfies additive analogues of the chain rule and Bayes rule. We farther develop the analogy to probabilities by describing a directed graphical representation that relies on our concept of conditional independence. One advantage of this model is that it leads to a natural structured elicitation process, very similar to that of Bayesian networks.

AB - Several schemes have been proposed for compactly representing multiattribute utility functions, yet none seems to achieve the level of success achieved by Bayesian and Markov models for probability distributions. In an attempt to bridge the gap, we propose a new representation for utility functions which follows its probabilistic analog to a greater extent. Starting from a simple definition of marginal utility by utilizing reference values, we define a notion of conditional utility which satisfies additive analogues of the chain rule and Bayes rule. We farther develop the analogy to probabilities by describing a directed graphical representation that relies on our concept of conditional independence. One advantage of this model is that it leads to a natural structured elicitation process, very similar to that of Bayesian networks.

UR - http://www.scopus.com/inward/record.url?scp=71549163440&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-04428-1_17

DO - 10.1007/978-3-642-04428-1_17

M3 - Conference contribution

AN - SCOPUS:71549163440

SN - 3642044271

SN - 9783642044274

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 192

EP - 202

BT - Algorithmic Decision Theory - First International Conference, ADT 2009, Proceedings

T2 - 1st International Conference on Algorithmic Decision Theory, ADT 2009

Y2 - 20 October 2009 through 23 October 2009

ER -