摘要
为了开发对科研实验数据的非线性关系精确建模工具,该文讨论了将遗传算法GA和神经网络算法结合,并使用Matlab软件实现计算过程。针对一个非线性测试函数逼近问题,设计了Matlab软件的GA算法的实现过程,并实验测试分析了GADS工具箱算子选择和参数设置。比较了单纯GA方法和GA结合Levenberg-Marquardt BP方法局部寻优的效果。结果表明实验中设计的基于Matlab的GA神经网络计算方案是一种有效的高精度模型,算法设计实现过程有指导意义,能为各领域提供有力复杂非线性建模工具。
In order to develop an exact non-linear relationship scientific experimental data modeling tool, the combination of genetic algo-rithm with artificial neural network and its realizations based on Matlab were discussed in this paper. To solve a nonlinear test function approximation problem, an implementation process of the GA algorithm based on Matlab was designed. GA operators selection and parameter settings of GADS toolbox were evaluated and analyzed. Comparison of pure GA algorithm with Levenberg-Marquardt BP Local optimization after GA method was made. The results demonstrate that the designed hybrid GA-NN method is an effective high-precision model which can be used in many areas for complex non-linear modeling.
出处
《电脑知识与技术》
2009年第12Z期10063-10065,共3页
Computer Knowledge and Technology
