1

所以我一直在使用 caret 包通过使用 train 函数来执行诸如 MLR、逐步回归和随机森林之类的东西。我这样做是为了让我也可以进行 10 次 10 折交叉验证，并且我通过执行以下操作来做到这一点：-

library(caret)
ctrl <- trainControl(method = "repeatedcv", number = "10", repeats = "10", savePred = T)
cadets.mlr <- train(RT..seconds~., data = cadets, method = 'lm', trControl = ctrl)

然后我可以从中查看预测值和观察值并将它们相互绘制等，以查看模型的准确度。RT..seconds 是我想针对 180 个不同实例（即 180 x 160 数据集）的 160 个其他变量建模的变量。

我想做同样的事情，但通过使用神经网络模型。我希望学习量为 0.25，动量为 50。我遇到了“AMORE”并尝试做同样的事情，但使用它来代替：-

cadets.nn <- train(RT..seconds.~., data = cadets, method = 'AMORE', trControl = ctrl)

但我不断收到以下消息：-

Error in modelLookup(method) : value of model unknown

我究竟做错了什么？当我通过插入符号调用 Amore 时，如何调整参数？

谢谢！

4

1 回答 1

1

AMORE isn't currently wrapped by train. It has been requested previously, but it is difficult to abstract AMORE's modeling code to work with the possible cases that people may want.

The neural network models that are wrapped by train can be found here. As of this writing there are 13 different neural network models available.

If none of those are what you would like and you are familiar with the AMORE package, you could write your own model code. See this page for instructions and examples on how to do that.

Max

于 2014-02-03T21:00:15.467 回答