Integration of artificial neural networks and knowledge bases

This article reviews the questions and directions of integration of artificial neural networks with knowledge bases. Two main directions of integration are considered: the inputs and outputs of artificial neural network to use integration of knowledge bases and artificial neural networks for solutions of application problems; by artificial neural network representation on the basis of ontological structures and its interpretation by means of knowledge processing in the knowledge base providing an intelligent environment for the development, training and integration of different artificial neural networks compatible with knowledge bases. The knowledge bases that are integrated with artificial neural networks are built on the basis of homogeneous semantic networks and multiagent approach to represent and process knowledge. An ontological model for representing artificial neural networks and their specifications within the framework of the model of unified semantic representation of knowledge is proposed. It is distinguished by the ability to represent artificial neural networks, its dynamics and other types of knowledge, including the specifications of artificial neural networks, as the common language for the representation of knowledge with a common theoretical-model semantics. A multiagent model for solving problems using artificial neural networks and other types of knowledge is proposed. It is distinguished by the interaction of agents in accordance with a given temporal model through a common memory that stores knowledge integrated into a single knowledge base.