|Date:||2021, May 25|
|Time:||10:00 a. m.|
|Title:||Inductive Logic Programming|
Logic programming (Prolog) uses universally quantified Horn clauses to represent knowledge. Program execution is deductive reasoning over such Horn clauses. Inductive logic programming learns Horn clauses from background knowledge and positive and negative examples. This is a form of inductive reasoning. Inductive logic programming is more flexible than machine learning approaches like neural networks because several predicates (of varying arity) can be learned at once, and because backgrund knowledge can be taken into account. The current interest in inductive logic programming is that it has been used as a base-line for neural-symbolic approaches, namely neural logic machines (which have been introduced in the Oberseminar on april 27) as well as differentiable logic machines (which actually can extract a logic program from examples).