How to remove NaN values from a given NumPy array?

Given a multidimensional NumPy array containing NaN (Not a Number) values, remove all NaN entries to get only the valid numerical data.

Example:

Input: [[5, nan, 8],
[2, 6, nan],
[nan, 1, 3]]

Output: [5. 8. 2. 6. 1. 3.]

Using ~np.isnan()

The ~ operator reverses the boolean array returned by np.isnan(), keeping only the non-NaN elements.

import numpy as np
arr = np.array([[12, 5, np.nan, 7],
                 [2, 61, 1, np.nan],
                 [np.nan, 1, np.nan, 5]])

res = arr[~np.isnan(arr)]
print("2D array converted to 1D after removing NaNs ->", res)

2D array converted to 1D after removing NaNs -> [12.  5.  7.  2. 61.  1.  1.  5.]

Explanation:

• np.isnan(arr): Creates a boolean array with True where values are NaN.
• ~np.isnan(arr): Inverts the boolean result so True indicates valid (non-NaN) values.
• arr[~np.isnan(arr)]: Selects and returns only the non-NaN elements.
• The result is flattened into a 1D array containing all valid numbers.

Using np.isfinite()

This method removes NaN and infinite values from a NumPy array. np.isfinite() returns True for all finite numbers, allowing you to keep only valid numeric elements.

import numpy as np
arr = np.array([[12, 5, np.nan, 7],
                 [2, 61, 1, np.nan],
                 [np.nan, 1, np.nan, 5]])

res = arr[np.isfinite(arr)]
print("2D array converted to 1D after removing NaNs ->", res)

2D array converted to 1D after removing NaNs -> [12.  5.  7.  2. 61.  1.  1.  5.]

Explanation: np.isfinite(arr): Returns True for all finite numbers (i.e., not NaN or Infinity).

Using numpy.logical_not() and numpy.isnan()

This method helps you filter out all NaN (Not a Number) values from a NumPy array. np.isnan() identifies the NaNs and np.logical_not() reverses the boolean result to select only the valid numbers.

import numpy as np

arr = np.array([[6, 2, np.nan], 
                 [2, 6, 1],
                 [np.nan, 1, np.nan]])
res = arr[np.logical_not(np.isnan(arr))]
print("2D array converted to 1D after removing NaNs ->", res)

2D array converted to 1D after removing NaNs -> [6. 2. 2. 6. 1. 1.]

Explanation:

• np.isnan(arr): Gives True for NaN elements.
• np.logical_not(...): Reverses that boolean mask, so True for valid numbers.

Related Articles

• How to Remove rows in Numpy array that contains non-numeric values
• Return the maximum of an array or maximum ignoring any NaNs