距离矩阵的计算

距离矩阵$D_{ij}^x={\Vert {\mathbf{x}}_i-{\mathbf{x}}_j\Vert }_2^2$的计算

给定数据矩阵 $\mathbf{X}\left(N\times D\right)$, 经常需要求解每两个样本间的距离$D_{ij}^x={\Vert {\mathbf{x}}_i-{\mathbf{x}}_j\Vert }_2^2$.

Remark: 这个距离矩阵的用处有很多，如核函数，对比学习, etc.

目前我所知有3种编程方法可以求解$D_{ij}^x$.

1. For Loop

import numpy as np

x = np.array([[1,5],[3,7],[7,9]])


def distance_forloop(x):
    N, _ = x.shape
    d = np.zeros((N,N))
    for i in range(N):
        for j in range(N):
            d[i,j] = np.sum((x[i,:]-x[j,:])*(x[i,:]-x[j,:]))
    
    return d

2. Elementary calculation

def distance_elementary(x):
    aa = bb = np.sum(x * x, axis = 1, keepdims = True)
    ab = np.matmul(x, x.T)
    return aa + bb.T - 2 * ab

3. Matrix Form

${\mathbf{D}}^x=diagnoal\left\{diag\left(\mathbf{X}{\mathbf{X}}^{\top }\right)\right\}11^{\top }+11^{\top }diagnoal\left\{diag\left(\mathbf{X}{\mathbf{X}}^{\top }\right)\right\}-2\mathbf{X}{\mathbf{X}}^{\top }$

where $diag\left(\mathbf{X}{\mathbf{X}}^{\top }\right)=\left[\Vert {\mathbf{x}}_1{\Vert }_2^2,\cdots ,\Vert {\mathbf{x}}_N{\Vert }_2^2\right]$.

def distance_matrixform(x):
    N, _ = x.shape
    diag = np.diagonal(np.matmul(x,x.T))# 这里得到的是向量
    diag = np.diag(diag)#还需要转换成对角矩阵
    one_one_t = np.ones((N,N))
    return np.matmul(diag,one_one_t) + np.matmul(one_one_t,diag) - 2*np.matmul(x, x.T)

Remakr: np.diagonal: (N,N)->(N,), np.diag:(N,)->(N,N)

经过检验这3种计算方式所得结果相同

d1 = distance_forloop(x)
print(d1)

d2 = distance_elementary(x)
print(d2)

d3 = distance_matrixform(x)
print(d3)

print((d1==d3)&(d1==d2))

但是方法2和方法3是类似的，计算效率都比1高，但是我个人倾向于方法3，因为它是公式化描述的。