Inductive Logic Programming (ILP)
Inductive Logic Programming (ILP) is a research area formed at the intersection of Machine Learning and
Logic Programming. ILP systems develop predicate descriptions from examples and background knowledge.
The examples, background knowledge and final descriptions are all described as logic programs and so, not very far
from natural language.
From a practical viewpoint, the user needs to describe his observations using a first order language,
then provide the file to the ILP machine which will extract the rules or the regularities. This is of course very similar
to a classical data mining or machine learning process. The main advantage of ILP is the fact that the results are
very similar to natural language sentences. You can get rules like
"if age(x,a) and (a < 40) and job(x, IT) then have_broadband(x)"
telling us that a IT person whose age is less than 40 has a broadband connection at home. Of course, these rules
are generally not sure 100% that is why sometimes we add to the rules a "probability" or "safety" factor measuring in some sense
the confidence we have in the rule.
From a theroretical viewpoint, ILP is based on proof theory and model theory for the first order predicate calculus.
The inductive process is characterised by techniques including
inverse resolution, relative least general generalisations, inverse implication and inverse entailment.
In BITE, we use a system developed in Japon by the team of Katsumi Inoue from
National Institute for Informatics.
>more on that subject
ILPNET is as well a good link to get more information about the current trends
in that field.
Your data will be processed by our ILP machine. You will receive the results by email soon.
|
