如何过滤出包含NaN的子数组​​?

问题描述

让我们假设一个形状为(n,5,2)的数组，该数组在随机位置包含NaN，由以下代码生成:

Let's assume an array of shape (n,5,2) which contains NaNs at random places, generated by the following code:

n = 10
arr = np.random.rand(n, 5, 2)

# replace some values by nan
arr = arr.ravel()
index_array = np.arange(arr.size)
np.random.shuffle(index_array)
arr[index_array[:5]] = np.nan
arr = arr.reshape(n, 5, 2)

如何有效过滤该数组，以便仅保留不包含NaN的那些arr[i]?然后将得到的形状为(m,5,2)和m<=n.

How can I efficiently filter this array such that only those arr[i]s are kept which don't contain NaNs? The resulting shape would then be (m,5,2) with m<=n.

推荐答案

无需重塑任何内容:

has_nans = np.isnan(arr).any(axis=(-1,-2))
has_nans
array([False, False, False,  True,  True,  True, False, False, False,  True], dtype=bool)

>>> arr = arr[~has_nans]
>>> arr.shape
(6, 5, 2)

较旧版本的numpy，您需要执行以下操作:

Older versions of numpy you will need to do something like the following:

has_nans = np.isnan(arr).any(axis=-1).any(axis=-1)

这篇关于如何过滤出包含NaN的子数组​​?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！