是否有更有效的方法来遍历矩阵并对特定列执行计算？

Question

我有一个包含 2 列的矩阵（矩阵 "X" 具有两个特征 - feature0 和 feature1）和可变行数。对于每个样本（矩阵中的行），我想计算一个扩展行，使得每一行都是 [feature0, feature1, feature0^2, feature1^2, feature0*feature1, 1].

我在下面编写了函数来完成这项工作。

def expand(X):

    X_expanded = np.zeros((X.shape[0], 6))

    for i in range(X_expanded.shape[0]):
        for j in range(X_expanded.shape[1]):

            if j <= 1:
                X_expanded[i, j] = X[i, j]
            elif j == 2:
                X_expanded[i, j] = X[i, 0]*X[i, 0]
            elif j == 3:
                X_expanded[i, j] = X[i, 1]*X[i, 1]
            elif j == 4:
                X_expanded[i, j] = X[i, 0]*X[i, 1]            
            elif j == 5:
                X_expanded[i, j] = 1



    return X_expanded

我的问题是，执行此计算是否更有效或 "better way"？对我来说似乎很麻烦，所以欢迎任何建议。提前致谢。

Answer 1

尝试制作一个简单的函数并将它们堆叠起来：

import numpy as np

def expanded(arr_2d):
    c1, c2 = arr.T
    return np.hstack([arr_2d, np.vstack([c1 ** 2, c2 ** 2, c1 * c2, np.ones(c1.shape[0])]).T])

大约快 145 倍：

arr = np.random.randint(0, 100, (10000, 2))

%timeit expand(arr)

# 41 ms ± 3.04 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)

%timeit expanded(arr)

# 282 µs ± 10.2 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

验证检查：

np.all(expand(arr) == expanded(arr))
# True

是否有更有效的方法来遍历矩阵并对特定列执行计算？

Is there a more efficient method to iterate over a matrix and perform calculations on specific columns?

python

iteration

loops

if-statement

matrix-multiplication