How to remove specific elements in a numpy array

Python meets immutability

In Numpy, arrays are immutable, meaning alterations spawn new arrays, like a cellular automaton. So perform operations wisely to save memory and uphold execution performance.

Index-based removal: np.delete to the rescue

To remove elements by index, use np.delete(). It's your loyal servant:

arr = np.array([1, 2, 3, 4, 5])
indices_to_remove = [1, 3]  # Indices of the unchosen ones

# Excommunicate elements at indices 1 and 3
new_arr = np.delete(arr, indices_to_remove)

print(new_arr)  # [1, 3, 5]


But remember, np.delete shows respect to the original array, and provides you with a fresh array.

The art of exclusion: employing set operations

np.setdiff1d() is the Picasso of subtracting numpy sets, it easily sketches out the elements to exclude:

arr = np.array([1, 2, 3, 4, 5])
exclude = np.array([2, 4])

new_arr = np.setdiff1d(arr, exclude)
print(new_arr)  # [1, 3, 5]


Shields up: strategic use of boolean masks

Boolean masks, like knights in shining armor, protect your array from unwanted elements. Here's how to command your knights using np.arange and np.isin:

arr = np.array([1, 2, 3, 4, 5])
mask = np.isin(np.arange(arr.size), [1, 3])

new_arr = arr[~mask]
print(new_arr)  # [1, 3, 5]


Masterclass in advanced removal techniques

Combining conditions: Powers unite!

Numpy gives you the superpower to combine conditions using logical operations. Create power-packed masks like:

arr = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9])
# Kick out even numbers and numbers greater than 7 (I mean, they had it coming right?)
new_arr = arr[~((arr % 2 == 0) | (arr > 7))]

print(new_arr)  # [1, 3, 5, 7]


Using np.where for sneak attacks

Suppose you want to remove elements based on a condition, like a predator stalking its prey. np.where() is your perfect ally:

arr = np.array([1, 2, 3, 4, 5])
# Get indices of elements cowering below 4
indices = np.where(arr < 4)

# Use np.delete() to remove these elements. It's almost like magic!
new_arr = np.delete(arr, indices)

print(new_arr)  # [4, 5]


Multi-directional assault on multidimensional arrays

For dealing with multidimensional arrays, consider axis parameters like a strategic genius:

arr = np.array([[1, 2], [3, 4], [5, 6]])
# Unseat the first and third row royals
new_arr = np.delete(arr, [0, 2], axis=0)

print(new_arr)  # [[3, 4]]
