Question

我正在尝试如下训练神经网络：

    # just an example 
    def func(): # define a function
       func.y = 4 # here y is a local variable, which I want to access; func.y defines 
                  # a method for my example function which will allow me to access 
                  # function's local variable y
       x = func.y + 8 # this is the main task for the function: what it should do
       return x

    func() # now I'm calling the function
    a = func.y # I put it's local variable into my new variable
    print(a) # and print my new variable

它工作正常，但我必须纠正df的偏斜，该偏斜具有99％的错误和1％的真实。 nnet（）函数有一个权重向量可以做到这一点，Neuronet（）是否有类似的东西？如果没有，我该如何解决？

Answer 1

我现在已经像这样修复了它，但是仍然对该软件包提供的任何工具感兴趣：

factor.list <- lapply(1:factor, function(x){df1[df1$a == 0,]})
factor.df <- do.call(rbind, factor.list)
model2 <- neuralnet(a ~ b + c + d + e ,data = rbind(df1,factor.df), hidden = c(10,10), lifesign = "full")

其中“因数”是1s和0s之间的比例

在Neuronet MLP R上使用重量

1 个答案: