Neural networks and relational database

Relation database systems are known of table processed data. Tables are presented by mathematical relations with attributes and domains. This allows to represent the real world objects (entities) and their mutual relations. On the other side, the neural networks are able to evaluate these objects by paralel processing of their attributes. Thus, basic characteristic of the neural networks can be understood as a method for nonprocedural data manipulations. In the article we present suggestion of using neural networks as processors able to sort or to classify and generally spaking to process database objects. The real objects are defined by means of relation tables, which records are changing in the real time. This is why using paralelism of neural networks on nonprocedural data processing is convenient. We suppose that attributes of relation table are identical with symptoms that are on the network inputs. This conception has been tested and verified using simple relation model.