Abstract The possibilities of signal binding in recurrent neural networks with controlled elements are investigated. It is shown that a variety of dynamic space–time structures with new associative properties can… Click to show full abstract
Abstract The possibilities of signal binding in recurrent neural networks with controlled elements are investigated. It is shown that a variety of dynamic space–time structures with new associative properties can be formed in the framework of such networks. A comparative analysis of the properties of linear, spiral single-level and multilevel structures of recurrent neural networks is carried out. Special attention is paid to the possibilities of controlling the associative-spatial interaction of signals in recurrent neural networks. The models of impulse neurons interaction are refined. The results of modeling of associative-spatial signal binding in two-layer recurrent neural networks with different logical structures of the layers are presented.
               
Click one of the above tabs to view related content.