libin毕业论文基于MATLAB神经网络仿真.docVIP

┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ 装 ┊ ┊ ┊ ┊ ┊ 订 ┊ ┊ ┊ ┊ ┊ 线 ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ 安徽工业大学管理科学与工程学院 基于MATLAB神经网络仿真 共 45 页 第 PAGE 4 页 摘 要 随着人工神经网络的研究和应用越来越广泛，误差反向传播算法（BP算法）的提出，成功地解决了求解非线性连续函数的多层前馈神经网络权值调整问题，BP神经网络如今成为最广泛使用的网络，研究它对探索非线性复杂问题具有重要意义，而且它具有广泛的应用前景。以BP神经网络为例，讨论了BP神经网络及几种改进BP神经网络性能的算法；通过BP学习算法的推导和分析得知BP网络是一种多层前馈网络，采用最小均方差的学习方式，缺点是仅为有导师训练，训练时间长，易限于局部极小；运用MATLAB来实现各种BP神经网络的实现的设计与训练，比较不同BP神经网络的性能，验证改进BP网络的优势，得出如何根据对象选取神经网络的结论。 关键词： 人工神经网络、BP神经网络、误差反向传播算法、MATLAB、仿真 Abstract With the artificial neural network of research and application of more and more widely, the error back-propagation algorithm (BP algorithm) is proposed, successfully resolved the continuous function for solving nonlinear multi-layer feed-forward neural network weights adjustment, BP network has become now the most widely used networks, Study to explore its complicated nonlinear problem has important significance, but also has broad application prospects. BP neural network is discussed and several improvements in the performance of BP neural network algorithm. BP learning algorithm through the derivation and analysis that the BP network is a multi-layer feedforward networks, the use of least-mean-variance approach to learning, there is only disadvantage is that the training instructors, training time, limited to local minimum easily. The use of MATLAB to achieve a variety of BP neural network to achieve the design and training, to compare the performance of BP neural network to verify the advantages of improving the BP network, how to draw the object selected in accordance with the conclusions of neural networks. Key words: Artificial neural network, BP neural networks, error back-propagation algorithm, MATLAB, simulation 目 录 TOC \o 1-3 \h \z \u HYPERLINK \l _Toc231979677 1.绪论 PAGEREF _Toc231979677 \h 5 HYPERLINK \l _Toc231979678 1.1 引言 PAGEREF _Toc231979678 \h 5 HYPERLINK \l _Toc231979681 1.2 神经网络概述 PAGEREF _Toc231979681 \h 5 HYPERLIN