神经网络学习历程4.docVIP

神经网络学习笔记4：CPN网络的实现 分类： 人工智能 2006-11-19 16:18 1791人阅读 评论(0) 收藏 举报 ??????? 对向传播网络（Counter Propagation)，简称CPN,是将Kohonen特征映射网络与Grossberg基本竞争型网络相结合，发挥各自长处的一种新型特征映射网络，被广泛的运用于模式分类，函数近似，数据压缩等方面。??????? CPN网络分为输入层，竞争层，隐含层。输入层与竞争层构成SOM网络，竞争层与输出层构成基本竞争 型网络，从整体上看，CPN网络属于有教师学习型网络，而由输入层和竞争层构成的SOM网络又属于典型的无教师网络，因此，这一网络既汲取了无教师型网络分类灵活，算法简练的特点，又采纳了有教师型网络分类精确的长处，使两种不同类型的网络结合起来。 ????? 至于CPN网络的学习算法，这里不打算多提，有兴趣的请参考相关书籍。这里给出一个简单的实现CPN网络的代码. ?程序实现功能：CPN神经网络?????作??? 者： laviewpbt???? 联系方式： laviewpbt@?? ? QQ?? ? 版本：Version 1.1.0???? 说明：复制请保留源作者信息，转载请说明，欢迎大家提出意见和建议 ? Private mW1() As Double???????????? 隐含层的权值?????????? S1? X? RPrivate mW2() As Double???????????? 输出层的权值?????????? S2? X? RPrivate mErr() As Double??????????? 误差Private mS1 As Long???????????????? 隐含层的神经元个数???? S1Private mS2 As Long???????????????? 输出层的神经元个数???? S2Private mR As Long????????????????? 输入层神经元个数?????? RPrivate mGoal As Double???????????? 收敛的精度Private mLr As Double?????????????? 隐含层学习速度Private mGama As Double???????????? 输出层学习系数Private mMaxEpochs As Long????????? 最大的迭代次数Private mIteration As Long????????? 实际的迭代次数 ****************************************? 中间变量?? ******************************************* Private HiddenOut() As Double?????? 输出层的输出Private OutCopy() As Double??? 比较的Private Ts As Long????????????????? 输入向量的总个数Private Initialized As Boolean????? 是否已初始化 ****************************************? 属性?? ******************************************* Public Property Get W1() As Double()??? W1 = mW1End Property Public Property Get W2() As Double()??? W2 = mW2End Property Public Property Get Err() As Double()??? Err = mErrEnd Property Public Property Get S1() As Long??? S1 = mS1End Property Public Property Let S1(Value As Long)??? mS1 = ValueEnd Property Public Property Get S2() As Long??? S2 = mS2End Property Public Property Get Goal() As Double??? Goal = mGoalEnd Property Public Property Let Goal(Value As Double)??? mGoal = ValueEnd Property Public Property Get Lr() As