How to Use numpy cumsum for Cumulative Sums

Use numpy.cumsum() to get the cumulative sum of array elements along a specified axis. It returns an array where each element is the sum of all previous elements plus the current one.

Syntax

The basic syntax of numpy.cumsum() is:

• numpy.cumsum(a, axis=None, dtype=None, out=None)

Where:

• a: Input array to compute the cumulative sum.
• axis: Axis along which the cumulative sum is computed. Default is None, which flattens the array.
• dtype: Data type of the returned array. If not specified, it uses the input array's dtype or a default type with higher precision.
• out: Optional output array to store the result.

python

numpy.cumsum(a, axis=None, dtype=None, out=None)

Example

This example shows how to use numpy.cumsum() on a 1D and 2D array to get cumulative sums along different axes.

python

import numpy as np

# 1D array example
arr1d = np.array([1, 2, 3, 4])
cumsum_1d = np.cumsum(arr1d)

# 2D array example
arr2d = np.array([[1, 2, 3], [4, 5, 6]])
cumsum_2d_axis0 = np.cumsum(arr2d, axis=0)  # sum down rows
cumsum_2d_axis1 = np.cumsum(arr2d, axis=1)  # sum across columns

print('1D cumulative sum:', cumsum_1d)
print('2D cumulative sum along axis 0:\n', cumsum_2d_axis0)
print('2D cumulative sum along axis 1:\n', cumsum_2d_axis1)

Output

1D cumulative sum: [ 1  3  6 10]
2D cumulative sum along axis 0:
 [[1 2 3]
 [5 7 9]]
2D cumulative sum along axis 1:
 [[ 1  3  6]
 [ 4  9 15]]

Common Pitfalls

Common mistakes when using numpy.cumsum() include:

• Not specifying the axis when working with multi-dimensional arrays, which causes the array to flatten and may give unexpected results.
• Assuming the original array is not changed; cumsum returns a new array and does not modify the input.
• Using integer dtype for large sums can cause overflow; specify a larger dtype like np.int64 if needed.

python

import numpy as np

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

# Wrong: forgetting axis flattens the array
wrong_cumsum = np.cumsum(arr)

# Right: specify axis to get cumulative sum along rows
right_cumsum = np.cumsum(arr, axis=1)

print('Wrong cumsum (flattened):', wrong_cumsum)
print('Right cumsum (axis=1):', right_cumsum)

Output

Wrong cumsum (flattened): [ 1  3  6 10]
Right cumsum (axis=1): [[1 3]
 [3 7]]

Quick Reference

Tips for using numpy.cumsum():

• Use axis=None to flatten the array before summing.
• Specify axis to sum along rows (axis=1) or columns (axis=0) in 2D arrays.
• Set dtype to avoid overflow for large sums.
• The function returns a new array; original data stays unchanged.

Key Takeaways

Use numpy.cumsum() to get cumulative sums of array elements easily.

Specify the axis parameter to control the direction of summation in multi-dimensional arrays.

The function returns a new array and does not modify the original array.

Set dtype to a larger type if you expect large sums to avoid overflow.

Without axis, the array is flattened before summing.