如何重塑 numpy 数组以删除所有带有 nan 的元素？

Question

I have a numpy array in python that is structured something like this:

array([[NaN, NaN, NaN],
       [a, b, NaN],
       [c, d, e],
       [NaN, NaN, f])

Each column has the same number of non-NaN values but they start and end at different rows. I would like to end up with an array structured like:

array([a, b, e],
      [c, d, f])

(That is, with all of the NaNs removed and the shape of the array modified to fit the number of data points.) However, I'm not sure how to do this efficiently.

I tried using:

x = x[~numpy.isnan(x)] 

and

x = x[numpy.logical_not(numpy.isnan(x))] 

but these both flattened the array.

Answer 1

转置数组，过滤掉 NaN，并重新整形为其原始形状（ 2是每列非 NaN 值的数量）。

x = x.T[~numpy.isnan(x.T)].reshape(-1, 2).T

Answer 2

Assuming what you're wanting to preserve is the number of columns:

num_cols = x.shape[1]
x_trimmed = x[~numpy.isnan(x)].reshape(-1, num_cols)

You shouldn't need to transpose the array.