Numpy索引越界报错求助：IndexError索引10超出轴1尺寸10

Hey there! Let's break down why you're hitting that IndexError and how to fix it.

What's Going Wrong? #

Your array Z has a shape of (500, 10)—that means it has 500 rows, each with 10 elements. Valid indices for the second axis (columns) only go from 0 to 9 here. But your y array is np.arange(500), which generates values from 0 all the way up to 499. When you use Z[range(500), y], you're trying to index columns with numbers like 10, 11, ..., 499—way beyond the 10-element limit of that axis. That's exactly why you get the error: index 10 is out of bounds for axis 1 with size 10.

How to Fix It #

You need to make sure every value in y is a valid index for the second axis of Z (i.e., between 0 and 9). Depending on your intended behavior, here are a couple common solutions:

1. Cycle through the 10 columns (most likely what you wanted)

If you want to pick one element from each row, cycling through the 10 columns repeatedly, use modulo to wrap y into the valid index range:

import numpy as np

Z = np.random.random((500,10))
y = np.arange(500) % 10  # Modulo 10 ensures values stay 0-9
print(Z.shape)  # Output: (500, 10)
print(y.shape)  # Output: (500,)

result = Z[range(500), y]
print(result.shape)  # Output: (500,) — one value per row!

2. Use valid indices only

If you meant to index each row with a random valid column index, generate y to only contain values between 0 and 9:

y = np.random.randint(0, 10, size=500)  # Generates 500 integers in 0-9 range

Key Takeaway for NumPy Advanced Indexing #

When using two arrays to index a 2D array (like Z[rows, cols]), every value in the cols array must be a valid index for the second axis of Z—meaning it has to be less than Z.shape[1] (which is 10 in your case).

内容的提问来源于stack exchange，提问作者Daniel Chepenko