BP算法的改进及其应用.doc

BP算法的改进及其应用 内容摘要 随着人工神经网络的发展，其用途日益广泛，应用领域也在不断扩展，已在人工智能、自动控制、计算机科学、信息处理、机器人、模式识别等各个领域中有着成功的案例。在众多神经网络中，又以BP（Back Propagation）网络的应用最为广泛，它所采用的BP算法已成为目前应用最为广泛的神经网络学习算法，绝大部分的神经网络模型都是采用BP算法或它的变化形式。这样的算法具有很好的非线性映射能力、泛化能力、容错能力，已在各个领域中取得了广泛的应用。但是人们在使用过程中却发现，这种算法存在这样或那样的局限，比如收敛速度慢、容易陷入局部最小值以及忘记旧样本的趋势，这些局域性严重影响了BP算法的应用。 本文主要针对BP算法的缺点，从梯度下降法和BP算法融合的角度进行改进，设计出效果较优的算法。主要研究工作如下： 第一、多层前馈模型的综述。 第二、BP算法的推导过程及其改进。 第三、实例仿真。 关键字：BP算法，神经网络，前馈模型. ABSTRACT With the development of artificial neural networks, it is applied widely in more fields, such as artificial intelligence, intelligent control, computer science, information processing, robotics, pattern recognition. BP (back propagation) neural network is one of the most widely applied neural networks. BP algorithm has become the most widely applied neural network algorithms, it and its deformations are used in most neural network models. These algorithms which have good nonlinear mapping ability, generalization ability and fault algorithm have wide applications in various engineering fields. However, the standard BP algorithm or its improved algorithms are based on steepest descent algorithm, thus there are some shortcomings, such as slow convergence, easy to fall into local minimum and forget the old samples. These shortcomings seriously affect the application of BP network. In this thesis, the shortcomings of BP algorithm is studied, from the perspective of combining gradient algorithm with BP algorithm, the optimum algorithms are designed. The main points of research are as follows: First of all, Summary of multilayer feed-forward model Secondly, Derivation of BP algorithm and its improvement Thirdly, The simulation Keywords: BP algorithm; neutral networks; feed-forward model; 目录 1 引言 4 2 基于BP算法的多层前馈模型 4 2.1数学模型 4 2.2各层计算 5 3 BP学习算法 6 3.1 网络误差与权值调整 6 3.2 算法推导 6 4 基于BP的多层前馈网的主要能力、局限性及改进 8 4.1 主要能力 8 4.2 局限性及改进 8 5.仿真实例 8 6.附录 10 7.参考文献...............................................................................................11 1