Question

我有一个矢量列表作为numpy数组。

[[ 1., 0., 0.],
 [ 0., 1., 2.] ...]

它们都具有相同的尺寸。如何在向量空间中找出哪个向量最接近数组中所有其他向量？有scipy或sklearn函数来计算吗？

Update：

“最接近”，我的意思是余弦和欧氏距离。

Update 2：

假设我有4个向量（a，b，c，d），向量之间的余弦距离为：

a，b = 0.2

a，c = 0.9

a，d = 0.7

b，c = 0.5

b，d = 0.75

c，d = 0.8

因此对于每一个，矢量a，b，c，d得到：

{
    'a': [1,0.2,0.9,0.7],

    'b': [0.2,1,0.5,0.75],

    'c' : [0.9,0.5,1,0.75],

    'd' : [0.7,0.75,0.8,1]
}

有没有办法说让我们说矢量d是与a，b，c最相似的那个？

Answer 1

你可以强行这样做。请注意，这是O（n ^ 2），对于大n来说会慢一些。

import numpy as np

def cost_function(v1, v2):
    """Returns the square of the distance between vectors v1 and v2."""
    diff = np.subtract(v1, v2)
    # You may want to take the square root here
    return np.dot(diff, diff)

n_vectors = 5
vectors = np.random.rand(n_vectors,3)

min_i = -1
min_cost = 0
for i in range (0, n_vectors):
    sum_cost = 0.0
    for j in range(0, n_vectors):
        sum_cost = sum_cost + cost_function(vectors[i,:],vectors[j,:])
    if min_i < 0 or min_cost > sum_cost:
        min_i = i
        min_cost = sum_cost
    print('{} at {}: {:.3f}'.format(i, vectors[i,:], sum_cost))
print('Lowest cost point is {} at {}: {:.3f}'.format(min_i, vectors[min_i,:], min_cost))

如何使用Python从最接近所有其他向量的向量列表中找到向量？

1 个答案: