NEUROCOMPUTATIONS IN RELATIONAL SYSTEMS

Artificial neural nets create a form of new high parallel computational structures used in many areas of applications. Fuzzy sets with all their conceptual capabilities and schemes of knowledge representation are considered as an interesting platform to cope with ambiguity present in human activity, especially decision processes. Relational structures in particular, forming a natural extension of boolean relational systems, play a significant role in building formal relational models of reality. In this correspondence we will indicate strong analogies between relational structures involving some composition operators and a certain class of neural networks. The problem of learning of connections of the structure is addressed and relevant learning procedures are proposed. An optimized performance index proposed here has a strong logical flavor. Some significant implementation details are studied as well.