#+!

<br />（四）multi-layer perceptron network<br /> <br />Can solve Exclusive Linear-non-classifiable input patterns，这解决了单层感知器的缺点.本质上是由于加了隐藏层的原因。<br /><br /> <br />Each layer equivalent to a Single-Layer Perceptron Networks<br />qth layer forms a nq-1 dimension

基本的NN概念1.Is a control method basically not depends on the Mathematical Model of operational objectives & environment2.Is suitable for the controlled objectives with uncertainty, complex, inaccuracy or high non-linearity3.Has strong Adapting, Study function

<br />BP网络其实还是前向反馈网络中的一种，她对于处理非线性分类问题有独特的优势。<br />BP NN can approach to any Non-Linear Mapping Relationship, only if there are enough hide layer & hide nodes.<br /> <br />Structure of BP Network is same as multi-Layer Perceptron Network only their Transfer

<br />Delta study principle:<br /> <br /><br /> <br /> <br />Hebbian study principle:<br /> <br /><br /> <br />Approach study principle:<br /> <br />

前面简单介绍了神经网络的概念，下面进一步了解。一。首先要清楚几个问题，什么时候可以使用？当系统要求不是严格100%输出正确率时。分类（classification)还是回归(regression)？通常状况下是分类，简单的线性分类或是非线性分类。回归是指使估计的曲线尽量接近所有分类点，实际上是分类的特殊情况。确定的(deterministic)还是随机的（stochastic)？weight权重可能有固定的最佳值，或是接近值，这是是确定的。随机的即是不确定的权重值。supervised还是unsupervi

上图是前馈网络基本结构，分别介绍几种前馈网络： M-P model，single-layer perceptron network,multi-layer perceptron network, BP(back propagation)(一）M-P modelProposed by McCulloch & PittsConsists of fixed Structure & fixed Link-WeightsRestrain Convex-Touch Weight ⇒ output is 0Excitin

<br />Design a DHNN (1)<br />1) Calculate Link-Weight Matrix<br />2) Obtain Attractor<br /> <br />Calculate Link-Weight Matrix:<br /> <br /><br /><br />not non negative defined symmetric.<br /><br /><br /> 

<br /><br /> <br /> <br /> <br /><br /><br /><br />Asynchronous Mode:<br /> <br /> <br /><br /><br /><br /><br /> <br /># Conclusion: from above calculation we can see<br />Using different adjusting sequence, x(5) can either weakly converge to x(2) <br />