This is particularly evident for the multilayer perceptron with 4 and 5 hidden units. The aggregate predictors perform better than the minimum only for 2 ...
確定! 回上一頁