Analysis And Design Of Some Classes Of Optimal Model Based On Neural Dynamics

In this paper, artificial neural networks of computational intelligence techniques, and linear algebraic equations, quadratic programming and nonlinear algebraic equation of three optimization problems were studied. From the large-scale network systems and real-time online control point of view, the optimization problems were corresponded to the neural network circuit based on the method of mathematical transformation. The designed neural network circuit model can generate the optimal solution online,which is suitable for very large scale integration.The design principles and methods that Hopfield neural network model to solve linear algebraic equations were studied. Its corresponding discrete systems were presented, and the stability and convergence of the discrete systems were discussed. Optimization model of neural network to solve quadratic programming with equality constrains were researched. By selecting an appropriate transformation of the variables in quadratic programming problems with equality constraints, a lower order recurrent neural network for solving higher quadratic programming is presented. The proposed recurrent neural network is globally exponential stability and converges to the optimal solutions of the higher quadratic programming. Especially for the case that the number of decision variables is close to the number of the constraints, the dimension of the current neural network is very low. A novel neural network model was also introduced in this paper,which can solve the problem of high order nonlinear algebraic equations. Some illustrative examples were discussed to show us how to design the analogue neural network.