Numpy数组元素减法操作后输出不符合预期的技术问题

Let's break down why your current code is returning a list of single-element numpy arrays instead of the 2D array you expect, and how to fix it step by step.

What's Going Wrong in the Original Code? #

Two key issues are causing the unexpected output:

1. Slicing returns an array instead of a scalar: When you use a[i, -1:], you're getting a 1D numpy array (e.g., array([6])) instead of a plain scalar value (like 6). Every calculation using this array will also produce a single-element array, which gets added to your k list.
2. Flat list instead of grouped rows: Your loop appends individual values/arrays directly to k, resulting in a flat list rather than a list of rows that can be converted into a 2D numpy array.

Solution 1: Use Numpy Vectorization (Most Efficient) #

Numpy is designed to avoid explicit loops for operations like this. We can use vectorized operations to compute the result in a clean, efficient way:

import numpy as np

a = np.array([[1,2,3,4,5,6],[50,51,52,40,20,30],[60,71,82,90,45,35]])

# Get the last element of each row, keeping column dimension for broadcasting
last_elements = a[:, -1:]
# Calculate absolute difference between last element and all preceding elements in the row
diff = np.abs(last_elements - a[:, :-1])
# Combine the differences with the original last elements to form the final array
k = np.hstack([diff, last_elements])

print(k)

Output: #

[[ 5  4  3  2  1  6]
 [20 21 22 10 10 30]
 [25 36 47 55 10 35]]

Solution 2: Modify the Original Loop #

If you prefer to keep the loop structure, you just need to adjust how you access the last element and group results by row:

import numpy as np

a = np.array([[1,2,3,4,5,6],[50,51,52,40,20,30],[60,71,82,90,45,35]])
k = []

for row in a:
    # Get the last element as a scalar (not an array) using [-1] instead of [-1:]
    last_val = row[-1]
    row_results = []
    # Calculate differences for all elements except the last one
    for num in row[:-1]:
        row_results.append(abs(last_val - num))
    # Add the original last element to the row
    row_results.append(last_val)
    # Add the completed row to k
    k.append(row_results)

# Convert the list of rows into a numpy array
k = np.array(k)

print(k)

This will produce the exact same 2D numpy array as the vectorized solution.

内容的提问来源于stack exchange，提问作者An student