在 Python 中，如何将 numpy 数组与矩阵的每一行进行比较以选择与向量最相似的行？

Question

For instance, if I have a 1D array [91, 119, 161, 203, 259] and a 2D array [[90,120,160,200,260], [95,115,165,204,255]] , how can I determine which row from the latter is a better match for the former?

In this case, it would be the first row because most numbers are only off by 1, whereas in the second row most numbers are off by 4. Also, the RMSE between the vector and the first row is 1.612, whereas for the second row it is 3.606.

Answer 1

There are many distance metrics, but cosine is a good measure for vector similarities. You can use scipy.spatial.distance.cosine to find the cosine distance. You'll want the vector with the smallest cosine distance.

In code:

import numpy as np
from scipy.spatial.distance import cosine

v = np.array([91, 119, 161, 203, 259])
matrix = np.array([[90,120,160,200,260], [95,115,165,204,255]])
assert np.argmin([cosine(v, row) for row in matrix]) == 0