信息与计算科学专业毕业论文--神经网络在数学建模中的应用.docVIP

摘 要 神经网络（Neural Network－NN,其迅猛发展将对整个信息科学产生巨大的影响。神经网络在数学建模中的应用也非常的广泛。 本文首先概述人工神经网络的基础，接着神经网络的模型、结构、。然后针对应用的目的，介绍了。SOM网，主要用来聚类和模式识别。 神经网络比较擅长的是作分类器，作预测，作参数选择，作控制器等等，这些问题都可以通过神经网络进行解决。 优点是计算灵活，尤其对稳态随机过程的逼近效果较好，，其一为函数逼近问题，其二为预测问题，如何结合实际应用问题来进行数据预处理、网络建立、网络训练及结果分析，展示BP神经网络在应用上的能力。最后，简要展望网络的发展前景Abstract The neural network (Neural Network-NN) is now the most valuable emerging discipline among newly-born science. It has developed into a new heat topic in modern science and technology, and its rapid development will have a huge influence on the entire information science. Neural network has also been widely applied to building up mathematic mold. The paper firstly outlines the basis of artificial neural network theory, and then elaborated on the model of neural networks, the structure, the classification and the properties. Then introduce the two commonly used mathematical modeling of neural networks according to the purpose of application. There are mainly two commonly used neural networks: one is based on the back-propagation algorithm for feed forward neural networks, that is, BP neural network, which is mainly used to achieve non-linear mapping; the other one is the self-organizing neural network, that is, SOM network, which is mainly used for clustering and pattern recognition. What make the neural network excels in mathematics modeling is its associative memory, and its ability to work as the sorter to make prediction and parameter choice, to work as the controller and so on. Through the neural network these functions can be easily work out. The application example is in the 2005 mathematics modeling competition, the usage of “the BP neural network approximation” to forecast the algorithm model, and to calculate the predicted value of different time periods in the observe station. “The BP neural network approximation” is the universal algorithm, the advantage is that its computation is flexible, especially when it is applied to the stable state stochastic pr