如何按出现次数移除两个numpy二维数组中的共同行？

核心思路 #

先统计两个数组中每行的出现次数，对同时存在于两个数组的行，取它们在两个数组中出现次数的最小值，从两个数组中各移除对应次数的该行。

实现代码 #

方法1：使用collections.Counter（直观易读）

import numpy as np
from collections import Counter

# 示例数组
A = np.array([[1, 2], [3, 4], [1, 2], [5, 6], [3, 4], [3, 4]])
B = np.array([[1, 2], [3, 4], [3, 4], [7, 8], [1, 2], [1, 2]])

# 将数组行转为可哈希的元组，用于统计次数
rows_A = [tuple(row) for row in A]
rows_B = [tuple(row) for row in B]

# 统计每行出现次数
count_A = Counter(rows_A)
count_B = Counter(rows_B)

# 找出共同行，并确定每个行需要移除的次数（取两者次数最小值）
common_rows = count_A.keys() & count_B.keys()
remove_counts = {row: min(count_A[row], count_B[row]) for row in common_rows}

# 过滤A数组
filtered_A = []
current_A_counts = count_A.copy()
for row in rows_A:
    if row in remove_counts:
        if current_A_counts[row] > remove_counts[row]:
            filtered_A.append(row)
            current_A_counts[row] -= 1
    else:
        filtered_A.append(row)
filtered_A = np.array(filtered_A)

# 过滤B数组
filtered_B = []
current_B_counts = count_B.copy()
for row in rows_B:
    if row in remove_counts:
        if current_B_counts[row] > remove_counts[row]:
            filtered_B.append(row)
            current_B_counts[row] -= 1
    else:
        filtered_B.append(row)
filtered_B = np.array(filtered_B)

# 输出结果
print("处理后的A数组：")
print(filtered_A)
print("\n处理后的B数组：")
print(filtered_B)

方法2：使用np.unique（更适合大型数组，性能更优）

import numpy as np

# 示例数组
A = np.array([[1, 2], [3, 4], [1, 2], [5, 6], [3, 4], [3, 4]])
B = np.array([[1, 2], [3, 4], [3, 4], [7, 8], [1, 2], [1, 2]])

# 统计A的唯一行及对应次数
unique_A, counts_A = np.unique(A, axis=0, return_counts=True)
count_A_dict = {tuple(row): cnt for row, cnt in zip(unique_A, counts_A)}

# 统计B的唯一行及对应次数
unique_B, counts_B = np.unique(B, axis=0, return_counts=True)
count_B_dict = {tuple(row): cnt for row, cnt in zip(unique_B, counts_B)}

# 确定共同行的移除次数
common_rows = count_A_dict.keys() & count_B_dict.keys()
remove_counts = {row: min(count_A_dict[row], count_B_dict[row]) for row in common_rows}

# 过滤A数组
filtered_A = []
temp_counts = count_A_dict.copy()
for row in A:
    row_tuple = tuple(row)
    if row_tuple in remove_counts:
        if temp_counts[row_tuple] > remove_counts[row_tuple]:
            filtered_A.append(row)
            temp_counts[row_tuple] -= 1
    else:
        filtered_A.append(row)
filtered_A = np.array(filtered_A)

# 过滤B数组
filtered_B = []
temp_counts = count_B_dict.copy()
for row in B:
    row_tuple = tuple(row)
    if row_tuple in remove_counts:
        if temp_counts[row_tuple] > remove_counts[row_tuple]:
            filtered_B.append(row)
            temp_counts[row_tuple] -= 1
    else:
        filtered_B.append(row)
filtered_B = np.array(filtered_B)

# 输出结果
print("处理后的A数组：")
print(filtered_A)
print("\n处理后的B数组：")
print(filtered_B)

结果说明 #

以示例数组为例：

• 原A数组中[1,2]出现2次，[3,4]出现3次；原B数组中[1,2]出现3次，[3,4]出现2次。
• 对共同行[1,2]，移除次数为2（取最小值），因此A中该行列全部被移除，B中保留1次；对[3,4]，移除次数为2，A中保留1次，B中该行列全部被移除。
• 最终处理后的数组会保留各自独有的行，以及共同行中“剩余次数”的行。

内容的提问来源于stack exchange，提问作者BGR