Probabilistic Lexicographic Preference Trees
Document Type
Conference Proceeding
Publication Date
1-1-2021
Abstract
We introduce probabilistic lexicographic preference trees (or PrLPTs for short). We show that they offer intuitive and often compact representations of non-deterministic qualitative preferences over alternatives in multi-attribute (or, combinatorial) binary domains. We specify how a PrLPT defines the probability that a given outcome has a given rank, and the probability that a given outcome is preferred to another one, and show how to compute these probabilities in polynomial time. We also show that computing outcomes that are optimal with the probability equal to or exceeding a given threshold for some classes of PrLP-trees is in P, but for some other classes the problem is NP-hard.
Publication Title
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume
13023 LNAI
First Page
86
Last Page
100
Digital Object Identifier (DOI)
10.1007/978-3-030-87756-9_6
ISSN
03029743
E-ISSN
16113349
ISBN
9783030877552
Citation Information
Xudong Liu and Miroslaw Truszczynski. 2021. Probabilistic Lexicographic Preference Trees. In Algorithmic Decision Theory: 7th International Conference, ADT 2021, Toulouse, France, November 3–5, 2021, Proceedings. Springer-Verlag, Berlin, Heidelberg, 86–100. https://doi.org/10.1007/978-3-030-87756-9_6