2D ID 数组和 1D 权重的加权 numpy bincount

2024-02-12

我使用 numpy_indexed 来应用矢量化 numpy bincount,如下所示:

import numpy as np
import numpy_indexed as npi
rowidx, colidx = np.indices(index_tri.shape)
(cols, rows), B = npi.count((index_tri.flatten(), rowidx.flatten()))

where index_tri是以下矩阵:

index_tri = np.array([[ 0,  0,  0,  7,  1,  3],
       [ 1,  2,  2,  9,  8,  9],
       [ 3,  1,  1,  4,  9,  1],
       [ 5,  6,  6, 10, 10, 10],
       [ 7,  8,  9,  4,  3,  3],
       [ 3,  8,  6,  3,  8,  6],
       [ 4,  3,  3,  7,  8,  9],
       [10, 10, 10,  5,  6,  6],
       [ 4,  9,  1,  3,  1,  1],
       [ 9,  8,  9,  1,  2,  2]])

然后我将分箱值映射到以下初始化矩阵的相应位置m:

m = np.zeros((10,11))
m 
array([[0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],
       [0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.]])

m[rows, cols] = B
m
array([[3., 1., 0., 1., 0., 0., 0., 1., 0., 0., 0.],
       [0., 1., 2., 0., 0., 0., 0., 0., 1., 2., 0.],
       [0., 3., 0., 1., 1., 0., 0., 0., 0., 1., 0.],
       [0., 0., 0., 0., 0., 1., 2., 0., 0., 0., 3.],
       [0., 0., 0., 2., 1., 0., 0., 1., 1., 1., 0.],
       [0., 0., 0., 2., 0., 0., 2., 0., 2., 0., 0.],
       [0., 0., 0., 2., 1., 0., 0., 1., 1., 1., 0.],
       [0., 0., 0., 0., 0., 1., 2., 0., 0., 0., 3.],
       [0., 3., 0., 1., 1., 0., 0., 0., 0., 1., 0.],
       [0., 1., 2., 0., 0., 0., 0., 0., 1., 2., 0.]])

然而,这考虑到每个值的权重index_tri每列是 1。现在,如果我有一个权重数组,请提供每列相应的权重值index_tri而不是 1:

weights = np.array([0.7, 0.8, 1.5, 0.6, 0.5, 1.9])

如何应用加权 bincount 以便我的输出矩阵m变成如下:

array([[3., 0.5, 0., 1.9, 0., 0., 0., 0.6, 0., 0., 0.],
       [0., 0.7, 2.3, 0., 0., 0., 0., 0., 0.5, 2.5, 0.],
       [0., 4.2, 0., 0.7, 0.6, 0., 0., 0., 0., 0.5, 0.],
       [0., 0., 0., 0., 0., 0.7, 2.3, 0., 0., 0., 3.],
       [0., 0., 0., 2.4, 0.6, 0., 0., 0.7, 0.8, 1.5, 0.],
       [0., 0., 0., 2.3, 0., 0., 2.4, 0., 1.3, 0., 0.],
       [0., 0., 0., 2.3, 0.7, 0., 0., 0.6, 0.5, 1.9, 0.],
       [0., 0., 0., 0., 0., 0.6, 2.4, 0., 0., 0., 3.],
       [0., 3.9, 0., 0.6, 0.7, 0., 0., 0., 0., 0.8, 0.],
       [0., 0.6, 2.4, 0., 0., 0., 0., 0., 0.8, 2.2, 0.]])

任何想法?


通过使用for循环和 numpybincount()我可以按如下方式解决:

for i in range(m.shape[0]):
   m[i, :] = np.bincount(index_tri[i, :], weights=weights, minlength=m.shape[1])

我正在尝试调整提供的矢量化解决方案here https://stackoverflow.com/a/60046212/1476932 and here https://stackoverflow.com/a/46256361/1476932分别但我不知道是什么ix2D变量对应于第一个链接中的变量。如果可能的话,有人可以详细说明一下吗?


更新(解决方案):

基于下面的 @Divakar 解决方案,这是一个更新版本,其中需要一个额外的输入参数,以防您的索引输入矩阵不覆盖输出初始化矩阵的整个范围:

    def bincount2D(id_ar_2D, weights_1D, sz=None):
        # Inputs : 2D id array, 1D weights array

        # Extent of bins per col
        if sz == None:
            n = id_ar_2D.max() + 1
            N = len(id_ar_2D)
        else:
            n = sz[1]
            N = sz[0]

        # add offsets to the original values to be used when we apply raveling later on
        id_ar_2D_offsetted = id_ar_2D + n * np.arange(N)[:, None]

        # Finally use bincount with those 2D bins as flattened and with
        # flattened b as weights. Reshaping is needed to add back into "a".
        ids = id_ar_2D_offsetted.ravel()
        W = np.tile(weights_1D, N)
        return np.bincount(ids, W, minlength=n * N).reshape(-1, n)

灵感来自this post https://stackoverflow.com/a/46256361/ -

def bincount2D(id_ar_2D, weights_1D):
    # Inputs : 2D id array, 1D weights array
    
    # Extent of bins per col
    n = id_ar_2D.max()+1
    
    N = len(id_ar_2D)
    id_ar_2D_offsetted = id_ar_2D + n*np.arange(N)[:,None]
    
    # Finally use bincount with those 2D bins as flattened and with
    # flattened b as weights. Reshaping is needed to add back into "a".
    ids = id_ar_2D_offsetted.ravel()
    W = np.tile(weights_1D,N)
    return np.bincount(ids, W, minlength=n*N).reshape(-1,n)

out = bincount2D(index_tri, weights)
本文内容由网友自发贡献,版权归原作者所有,本站不承担相应法律责任。如您发现有涉嫌抄袭侵权的内容,请联系:hwhale#tublm.com(使用前将#替换为@)

2D ID 数组和 1D 权重的加权 numpy bincount 的相关文章

随机推荐