如何根据指定索引过滤形状为(50000,32,32,3)的n维数组A生成新数组B

Hey there! Since your array A has a shape of (50000, 32, 32, 3), it’s clear the first dimension represents individual samples (each being a 32x32 RGB image). Grabbing only the samples at your specified indices is super straightforward with NumPy’s advanced indexing.

Here’s how you do it:

• First, make sure your target indices are stored as a NumPy array (a regular Python list works too). Let’s call this index collection target_indices (like the array you provided: array([5, 10, 15, ..., 49988])).
• Then, simply index your original array A using this index array:

import numpy as np

# Assuming A is your original array and target_indices is your index array
B = A[target_indices]

This will create a new array B where each element corresponds to the slice of A at the indices in target_indices. The shape of B will be (len(target_indices), 32, 32, 3)—exactly what you need!

A quick note: This works because NumPy allows integer array indexing, which lets you pick out specific elements along a dimension. If you ever needed to index a different dimension (say, the second one), you’d adjust the indexing to A[:, target_indices], but based on your question, the first dimension is clearly the one you’re targeting.

内容的提问来源于stack exchange，提问作者Manasi