Default logic - HiIAmTzeKean/SC3000-Artificial-Intelligence GitHub Wiki


tags:

  • 🌱
  • AI
  • ComputerScience date: 23--Apr--2023

Default logic

$$\frac{prerequisite:justification}{conclusion}$$

  • If prerequisite is true
    • All justifications are consistent
      • KB does not entail $\lnot beta(x)$
    • Conclusion is true

Semantics

  • $\Delta$ is set of default rules
  • $\Phi$ set of facts

Default theory

  • Closed world assumption
    • What is not known to be true is assumed false
    • $\frac{:\lnot \beta}{\lnot \beta}$
  • Default rules in $\Delta$ can be applied to a theory
    • If preconditions are entailed
    • And if justification are consistent
  • Other default rules can be applied to a resulting theory
    • To extend $\Phi$ known facts
  • An extension of a default theory
    • A theory has exhausted all default rules
    • A default theory can have 0, 1, or many extensions

Reiter inference

  • Beliefs $\Xi = \Phi$
  • Guess $\exists \Xi^* = \Xi \cup \alpha(c)$
    • For every $\delta \in \Delta$
      • Check triggered and justified
      • Update beliefs if consistent
      • $\Xi = \Xi \cup \gamma(c)$
  • Check if guess was valid

Nixon diamond

Default rules applied in different order can result in different extensions

  • Sceptical
    • Formula is entailed by default theory if it is entailed by all extensions
  • Credulous
    • Formula is entailed if it is entailed by at least one of its extensions
    • Can give rise to inconsistent belief depending on expansion order of default rule

Makinson approach

  • Each

Links: Semantic - Logic - NTU Vault/Machine Learning/Inference

⚠️ **GitHub.com Fallback** ⚠️