Distributed Decision Making and Plan Recognition Under Uncertainty
Abstract
Our new work on abstraction approximations developed techniques for enhancing qualitative and numeric reasoning. First we developed incremental techniques for a task we call tradeoff resolution, in the context of qualitative Bayesian networks Liu and Wellman 1998. One approach incrementally marginalizes nodes that contribute to the ambiguous qualitative relationships. Another approach evaluates approximate networks for bounds of probability distributions, and uses these bounds to determine the qualitative relationships in question. This approach is incremental in that the algorithm refined the state spaces of random variables for tighter bounds until the qualitative relationships are determined. Both approaches provide systematic methods for tradeoff resolution at potentially lower computational cost than application of purely numeric methods. Second, we developed techniques that exploit qualitative probabilistic relationships among variables for computing bounds of conditional probability distributions of interest in Bayesian networks Liu and Wellman 1998. Using the signs of qualitative relationships, we define abstraction operations that are guaranteed to bound the distributions of interest in the desired direction. By evaluating incrementally improved approximate networks, our algorithm obtains monotonically tightening bounds that converge to exact distributions. For some classes of utility functions, the tightening bounds monotonically reduce the set of admissible decision alternatives as well.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2000
- Accession Number
- ADA405436
Entities
People
- Michael P. Wellman
Organizations
- University of Michigan