文件名称:GMDH
介绍说明－－下载内容均来自于网络，请自行研究使用
gmdh代码 训练方法
——————————————
1.将数据集分为训练集和测试集。
2.建立输入层
3.建立所有符合5楼条件的神经元（只考虑“攻”和“受”的关系，不考虑具体的权值）。如果前面几层共有N个神经元，那么我们就需要建立N(N-1)/2个神经元。
4.用最小二乘法确定权值
5.计算每一个神经元在测试集上的表现。
6.选取表现最好的K个神经元放在后面一层。（K可以自己选择）
7.重复3,直到网络的性能足够好，或者网络开始过拟合。(GMDH codeTraining methods
--------------
1. Divide the data set into training sets and test sets.
2. Create an input layer
3. Establish all neurons that meet the conditions of the 5th floor (only consider the "attack" and "by" the relationship, regardless of the specific weight). If there are N neurons in the first few layers, then we need to establish N (N-1) / 2 neurons.
4. Determine the weight using the least squares method
5. Calculate the performance of each neuron on the test set.
6. Select the best performing K neurons on the back. (K can choose their own)
7. Repeat 3, until the network performance is good enough, or the network began to fit.)
——————————————
1.将数据集分为训练集和测试集。
2.建立输入层
3.建立所有符合5楼条件的神经元（只考虑“攻”和“受”的关系，不考虑具体的权值）。如果前面几层共有N个神经元，那么我们就需要建立N(N-1)/2个神经元。
4.用最小二乘法确定权值
5.计算每一个神经元在测试集上的表现。
6.选取表现最好的K个神经元放在后面一层。（K可以自己选择）
7.重复3,直到网络的性能足够好，或者网络开始过拟合。(GMDH codeTraining methods
--------------
1. Divide the data set into training sets and test sets.
2. Create an input layer
3. Establish all neurons that meet the conditions of the 5th floor (only consider the "attack" and "by" the relationship, regardless of the specific weight). If there are N neurons in the first few layers, then we need to establish N (N-1) / 2 neurons.
4. Determine the weight using the least squares method
5. Calculate the performance of each neuron on the test set.
6. Select the best performing K neurons on the back. (K can choose their own)
7. Repeat 3, until the network performance is good enough, or the network began to fit.)
相关搜索： gmdh
(系统自动生成,下载前可以参看下载内容)
下载文件列表
GMDH - 副本.txt