Reasoning under Uncertainty
1. The Problem
commonsense reasoning
to represent and process various types of uncertainty
focused on the uncertainty in truth value: revisible and multi-valued
required changes in the major components
2. Proposed Solutions
non-monotonic logic, e.g. default logic
Truth Maintenance Systems
and belief revision
fuzzy logic
probabilistic logic
and Bayesian network
Dempster-Shafer theory
and imprecise probability
3. Issues
choice among multiple extensions and reference classes, e.g. Nixon diamond
interpretation of the numerical values: probability and membership
truth-value revision in general
computational complexity
4. Readings
Sections 10.7-8, 13.1-6, 14.1-2, 14.7
5. Ideas
Uncertainty in Artificial Intelligence is an important topic; uncertainty should be represented and processed
reasonable:
- different types of uncertainty
- qualitative and quantitative representation
problematic:
- the origin and nature of uncertainty
analysis: