Neural Network And Artifical Fish-Swarm Algorithm For (Parametric) Programming With Linear Constraints

Linear programming is an important branch of operational research and scientific mathematics method. It is used widely in many fields, such as planning, organization and management analysis and decision-making in industry and agriculture and national defense construction. The theory of optical problem has been come of age since simplex method of linear programming was put forward in 1947. But the future development in this field is still very prospective. Real-time optimization is often needed in the science and engineering, such as robot control. Because neural network has the characteristics of cosmically collateral calculating and fast convergence, many researchers are trying to develop the neural network model of optical problem. Other intelligence computational methods are also hotspots.In this thesis, the linear constrained programming problems are discussed. A neural network algorithm is proposed which is based on the article of Zhang Guoping for solving the linear constrained programming problems. The algorithm is based on gradient and object function and restraint condition are differentiable. The constrained conditions are treated by penalty function. The optimization of linear constrained programming problems is solved. The algorithm proves effective and correct through some examples. Artificial fish-swarm algorithm for solving constrained optimization problems is also used. The constrained condition and object function are separate. And non-feasible solution has a certain proportion. At last, compared with neural network, artificial fish-swarm algorithm has a strong robustness and self-adaptation.