In this paper, a kind of dynamic genetic algorithm based on the neural network with constraints is presented, which combines the local searching ability of neural network with the global searching ability of genetic algorithm. The dynamic algorithm is used to decide the initial point of the neural network, and the neural network is employed to decide the fitness of the dynamic genetic algorithm. The proposed algorithm has some theoretical and biological meanings. Compared with the standard genetic algorithm, it can decrease the searching scale and get the approximate solution of quadratic programming problems that are non-definite.