Inductive Logic Programming
Inductive Logic Programming (ILP) (De Raedt 2008; Nienhuys-Cheng and De Wolf 1997) is a family of methods for automated learning (or machine learning) of general rules from specific data and background knowledge. Unlike other machine learning methods, ILP uses the expressive language of the first-order predicate logic to represent input data, background knowledge, and learned hypotheses. This makes ILP suitable for data mining applications in domains characterized by nontrivially structured data, such as biochemistry or natural language processing. Since learned hypotheses can acquire the form of logic programs, the goal of ILP may be formulated as automated induction of the latter; hence, the name inductive logic programming.
Consider a task where an ILP algorithm receives examples of toxic and nontoxic chemical compounds. From these examples, it learns a general hypothesis, according to which toxicity...
The author is supported by the project 103/10/1875 of the Czech Science Foundation.