模式识别课件 第8章.pptVIP

8.6.2 Neuron Model Network Architecture The architecture of a network consists of a description of how many layers a network has, the number of neurons in each layer, each layer’s transfer function, and how the layers connect to each other. The best architecture to use depends on the type of problem to be represented by the network. Any reasonable function can be represented with a two-layer network: a sigmoid layer feeding a linear output layer. Networks with biases can represent relationships between inputs and outputs more easily than networks without biases. Toolbox 8.6.3 Perceptrons Introduction Neuron Model Perceptron Architecture Learning Rules: perceptron learning rule (learnp) Training Limitations and Cautions Graphical User Interface Summary nnd4db/ nnd4pr ：newp Toolbox 8.6.4 Linear Filters Introduction Neuron Model Perceptron Architecture Learning Rules: LMS Algorithm, Widrow-Hoff learning algorithm (learnwh) Training Limitations and Cautions Summary ：newlin Toolbox 8.6.5 Backpropagation Introduction Model Architecture Training: backpropagation algorithms Faster Training Limitations and Cautions Summary : newff, newrb, newpnn Toolbox 8.6.6 Competitive Network Introduction Competitive Learning Self-Organizing Maps Learning Vector Quantization Networks Summary : newc, newsom, newlvq Toolbox 讨论 人工神经网络是由大量神经元经广泛互联组成的非线性网络，单元间的连接强度从样本中学习 神经网络模型决定于神经元类型，神经元间的连接方式和学习算法 MLP、RBF和SOFM是模式识别中应用最多的几种模型 习题 解释Hebb学习规则的基本思想 浏览Matlab的neural network toolbox的帮助文档（perceptron, linear filter, bp) 写出C-Means算法双层神经网络实现的算法描述或画出流程图 以两类情况为例，证明当神经网络输出采用“c中取1”编码，隐层和输出层神经元均采用Sigmoid函数，训练准则函数为最小MSE，则MLP的输出就是对贝叶斯后验概率的估计 感知器学习算法举例 前馈网络 感知器学习算法演示见nnd 3. 多层感知器 多层感知器: Multi-Layer Perceptron, MLP Architecture： 前馈网络 多层感知器的一致逼近性 逻辑功能：单个阈值神经元可以实现任意多输入的与、或及与非、或非逻辑门 任何逻辑函数可由两层前馈网络实现 当神经元的输出函数为Sigmoid等函数时，两层前馈网络可以逼近任意的多元非线性函数 MLP的适用范围大大超过单层网络 前馈网络 XOR问题 任何一个逻辑电路都可以只用XOR门来实现， XOR是通用门 (universal logic gate) 单层感知器不能解决XOR问题 两层感知器可以解决XOR问题 前馈网络 4. C-Means算法双层神经网络实现 采用双层神经网络结构，如图所示： 输入节点数等于输入模式矢量