Topology Recognition Of Multi-weight Coupled Neural Network

In recent years,the dynamical behaviors have been widely concerned in coupled neural networks(CNNs)by numerous researchers owing to their extensive applications in various fields including secure communication,image encryption,and so on.Particularly,the synchronization problem for CNNs has attracted widespread attention.However,most existing works about synchronization required that the CNNs has single weight.Considering that various influence factors have an effect on the dynamical behavior of CNNs,the CNNs may be more accurately described by multi-weighted network model.Therefore,it is challenging and meaningful to further investigate the synchronization for multiple weighted CNNs,but very few results have been derived about this topic.The coupling weights represent the topology of complex network.But,most interesting results about the synchronization require that the complex networks have precisely known topologies.Considering that the coupling weights of complex networks may be unknown in some occasions,many investigators have been increasingly interested in studying the topology identification for complex networks.Considering the aforementioned reasons,we investigate the topology identification problem for two classes of multiple weighted CNNs in this paper.The main contents and contributions are summarized as follows.This paper concentrates on the research of topology identification problem for CNNs with multiple state couplings and multiple delay state couplings,in which single neural network may or may not include time delay.By constructing the response networks,designing suitable controllers and parameter adjustment schemes,utilizing several inequality techniques and significant lemmas,the topology of the original networks are identified based on synchronization between the original networks and the response networks.Finally,the validity of the proposed controllers and parameter adjustment strategies is demonstrated by four numerical examples.