Dynamics Behaviors For Three Kinds Of BAM Neural Networks And The Application For SOFM Neural Networks

In this paper, the dynamics behaviors for three kinds of bi-directional associativememory neural networks and the application for self-organizing feature maps arestudied. Specifically, this paper covers the following five aspects:1、This paper briefly outline the significance, background, progress of the neuralnetwork and the achievements on this.2、The dynamics behaviors of Cohen-Grossberg type bi-directional associativememory neural network with S-type delays is studied. Firstly, by employing M-matrix,the property of homeomorphism and the Lyapunov function methods, the theorems ofthe existence and asymptotical stability of equilibrium point are obtained, it extendsthe results of the relevant literature. Secondly, by using the contraction mappingprinciple and Lyapunov function method, the existence and exponential stability ofequilibrium point are obtained.3、The stochastic factors are added to the Cohen-Grossberg type bi-directionalassociative memory neural network with S-type delays. Without assuming thesmoothness, by constructing suitable Lyapunov function, employing thesemi-martingale convergence theorem and the technique of some inequalities, somesufficient criteria to check the almost exponential stability of the model are obtained.4、The bi-directional associative memory neural networks with discrete anddistributed delays and impulses is studied. The exponential stability of equilibriumpoint is obtained via Lyapunov function methods and the technique of someinequalities. It extends the results of the relevant literature.5、The improved self-organizing feature maps (SOFM) network is proposed,which can map high-dimentional data into simple geometric relationships on alow-dimentional display effectively. Then use it on comprehensive evaluation ofstudents.