0

How can I infer rules with incomplete facts as input?

For example, I only has X eat flies as input: - {cf(croaks) 1.0}, {cf(eats flies) 1.0}, {cf(a frog) 1.0} - If X 'croaks' and X 'eats flies' Then X is 'a frog' {cf 1.0}

  • If X is 'a frog' Then X is 'green' {cf 1.0}

How can I infer X is green with some percentage accuracy?

MiP
  • 499
  • 1
  • 7
  • 16
  • In general case, you can't. What if the initial set contains *dragonflys* (they eat flys as well) only? In this case the `probability == 0`. What if the initial set has *frogs* only? Here we have the `probability == 1`. You have to have some ideas about distribution, probablility etc. to use, say, Bayesian inference – Dmitry Bychenko Oct 27 '16 at 13:37
  • @DmitryBychenko Could you show me other inferences? I cannot use Bayesian inference in this case because my data rules cannot be trusted. – MiP Oct 27 '16 at 13:48
  • even if you don't trust the rules, you can ascribe some probabilities (croacking fly-eater animal has 85% chances to be a frog; 10% of the animals in the ecosystem are fly-eaters etc.) and then perform Bayesian inference. – Dmitry Bychenko Oct 27 '16 at 13:57
  • From a document I read, Bayesian inference is said to be used only when collected data is reliable though. Thank you. – MiP Oct 27 '16 at 14:00

0 Answers0