在 Numpy 中的阵列上应用掩码

Question

I have a problem I can't figure out and I hope you can help me. Assume I have an array with the shape of 146,243, I have about 144 of therm. (12 months for 12 years) There are certain indexes which always should have np.nan values and I want them always to contain nan-Values, but I have no idea how. I have come thus far. (This example has the right shape but there are just two occurences for np.nan, I have much more in my actual array.)

import numpy as np

rand_array = np.random.rand(146,243)
# In every new array those positions should always be np.nan
rand_array[0][90] = np.nan
rand_array[12][15] = np.nan
# Searching for nan-Values
nan_vals = np.argwhere(np.isnan(rand_array))

# Next Iteration
rand_array_2 = np.random.rand(146,243)
# Apply the nan values at the position found by nan_vals to rand_array_2 at the same positions

This example is quite easy and I could just manually set the nan.values but my problem array has far too many values for this method to be feasable. How would I go on about from finding out once where the np.nan Values are, to applying them to every consecutive array? Since I have the position of them saved in nan_vals I just need to figure out how to apply it to another array.

Kind regards.

Answer 1

Numpy provides very much feasibilty.

You could use:

rand_array_2[nan_vals[:, 0], nan_vals[:, 1]] = np.nan

This will set (for your given example) rand_array_2[0, 90] and rand_array_2[12, 15] as nan