Interval-Valued Neural Multi-adjoint Logic Programs

Abstract

The framework of multi-adjoint logic programming has shown to cover a number of approaches to reason under uncertainty, imprecise data or incomplete information. In previous works, we have presented a neural implementation of its fix-point semantics for a signature in which conjunctors are built as an ordinal sum of a finite family of basic conjunctors (Gödel and Lukasiewicz t-norms). Taking into account that a number of approaches to reasoning under uncertainty consider the set of subintervals of the unit interval as the underlying lattice of truth-values, in this paper we pursue an extension of the previous approach in order to accomodate calculation with truth-intervals.