RBF网络对函数回归技术的研究.docVIP

RBF网络对函数回归技术的研究 摘要 人工神经网络具有大规模并行处理能力、分布式存储能力、自适应(学习)能力等特征，神经网络特有的非线性适应性信息处理能力。克服了传统人工智能方法的缺陷，已广泛应用于模式识别、信号处理等各种应用领域。MATLAB集数值分析、矩阵运算、信号处理和图形显示于一体，构成了一个方便的、界面友好的用户环境，应用MATLAB语言构造典型神经网络的激活传递函数。编写各种网络设计与训练的子程序．网络设计者可以根据需要调用工具箱中有关神经网络的设计训练的函数，从繁琐的编程中解脱出来。 本文介绍了人工网络结构和RBF神经网络的基本原理及主要特点，还对MATLAB进行系统介绍。重点阐述了利用MATLAB实现利用RBF网络对给定函数进行回归分析。并举例说明了基于MATLAB神经网络工具箱建立RBF神经网络模型机实现仿真的方法。给出MATLAB程序，对函数回归的可视化结果，并与原函数进行比较，分析差别的基本原因。 关键词：网络设计，回归分析，仿真，MATLAB神经网络工具箱，RBF神经网络 RBF network function regression Technology Abstract: Artificial neural networks have large-scale parallel processing, distributed storage capacity, adaptive (learning) ability characteristics, non-linear adaptive neural network-specific information processing capabilities. Overcome the shortcomings of traditional artificial intelligence has been widely used in pattern recognition, signal processing and other applications. MATLAB sets numerical analysis, matrix computation, signal processing and graphical display on the whole, constitute a convenient, user-friendly user environment, application of MATLAB language structures typical neural network activation transfer function. The preparation of a variety of network design and training routines. Web designers can call the neural network toolbox in the design of the training function, from the tedious programming freed. This paper introduces the artificial neural network RBF network structure and the basic principles and main features, but also a systematic description of the MATLAB. Focuses on the use of RBF network using MATLAB to achieve a given function by regression analysis. And examples of neural network toolbox of MATLAB-based RBF neural million in Romania to establish model aircraft to achieve simulation method. Matlab program is given on the return of visual function, the results were compared with the original function to analyze the underlying causes of differences. Key words：Network design ，Regression Analysis ，Simulation，MATLAB neural network toolbox，RBF ne