从给定的最小和最大先验创建一个 numpy 矩阵

Question

我有 3 个先验，给出了最小和最大范围。 通过使用它们，我需要以以下形式创建一个 NumPy 数组；

M = [[x_0, y_0, z_0], [x_1, y_1, z_1], ...,[x_N, y_N, z_N]]

其中x=[0.60, 0.80] , y=[1, 80] , z=[0.022, 0.024]我怎样才能以最有效的方式（即通过最少的代码和使用 NumPy）做到这一点？

Answer 1

import numpy as np

N = 4
xs = np.random.uniform(x[0], x[1], size=N)
ys = np.random.uniform(y[0], y[1], size=N)
zs = np.random.uniform(z[0], z[1], size=N)

res = np.vstack((xs,ys,zs)).transpose()

数字示例

N = 4
xs = np.random.uniform(0.6, 1.1, size=N)
ys = np.random.uniform(3, 5, size=N)
zs = np.random.uniform(8, 9, size=N)
res = np.vstack((xs,ys,zs))
res.transpose()
# array([[0.88860867, 3.11233047, 8.26189772],
#        [0.70096631, 4.984737  , 8.01999442],
#        [1.08111807, 3.54934757, 8.28137655],
#        [1.04116942, 3.16903737, 8.12647381]])

Answer 2

它可以通过以下方式实现：

np.array([x, y, z]).T

一些基准测试：

size = 3 * 1000
50 loops, best of 5: 8.14 µs per loop     # np.vstack
50 loops, best of 5: 2.95 µs per loop     # this answer

size = 3 * 10000
50 loops, best of 5: 27.3 µs per loop
50 loops, best of 5: 19.9 µs per loop

size = 3 * 100000
50 loops, best of 5: 383 µs per loop
50 loops, best of 5: 359 µs per loop

size = 3 * 1000000
50 loops, best of 5: 5.23 ms per loop
50 loops, best of 5: 5.09 ms per loop