返回非NA值numpy数组的所有行索引

I have a numpy array with a lot of NaN values and in some rows non-NA values.

我有一个numpy数组,有很多NaN值,在某些行中有非NA值。

Now, I want to get all indices of the rows that have a non-NA value (i.e. a numeric values) and want to extract the values for these specific row indices from a different numpy array. So we have:

现在,我想得到具有非NA值(即数值)的行的所有索引,并希望从不同的numpy数组中提取这些特定行索引的值。所以我们有:

Y_test, a np array of length 698 with on some rows numeric values

Y_test,一个长度为698的np数组,带有一些行数值

Y_test
array([ nan,  13.,  nan,  10.,  nan])

Y_pred, a np array of length 698 with numeric values on each row

Y_pred,一个长度为698的np数组,每行都有数值

Y_pred
array([ 0.96,  0.57,  0.33,  0.43,  0.83])

I want to only extract the numeric values of np_array2 on the indices of the rows with numeric values in np_array1.

我想只在np_array1中使用数值的行的索引提取np_array2的数值。

For example:

array([ 0.57, 0.43])

I am doing this to calculate the error between the non-NA rows from Y_test and the corresponding rows of Y_pred.

我这样做是为了计算Y_test的非NA行和Y_pred的相应行之间的误差。

Can anyone help me?

谁能帮我?

1 个解决方案

#1

You can use isnan to find the indices where an array is nan and then just use the inverse.

您可以使用isnan查找数组为nan的索引,然后使用反向。

np.isnan(Y_test)
array([ True, False,  True, False,  True], dtype=bool)
Y_pred[~np.isnan(Y_test)]
array([ 0.57,  0.43])

EDIT

Per further input I understand that your two arrays have different sizes. In that case you can use flatnonzero as @PaulPanzer pointed out.

根据进一步的输入,我知道你的两个数组有不同的大小。在这种情况下,您可以使用@nonulzero指出@PaulPanzer。

np.flatnonzero(~np.isnan(Y_test))
array([1, 3])
Y_pred[np.flatnonzero(~np.isnan(Y_test))]

#1