在Python中连接坐标数组

Question

我正在从多个d维立方体构建一个坐标数组。我想从一个空数组开始，并附加在每个for循环中生成的坐标。但是，np.concatenate（和vstack）需要一个匹配维度的现有数组。我还希望保留维度，而不是获得平面数组

这就是我想做的想法：

#sample from d-dimensional cube
def make_cubes(d, npoints, ncubes):
    cubes = []
    for i in range(ncubes):
        c = np.random.rand(npoints, d)     #sample from cube in d dimensions
        c += np.random.uniform(-5,5)   #random translation
        cubes = np.concatenate((cubes, c))
    return cubes

为了使它实际工作，我必须将多维数据集定义为

cubes = [[]]
for d in range(d):
    cubes[0].append(np.nan)

Answer 1

正如hpaulj所述，您可以使用append而不是concatenate。最后，可以将阵列重塑为所需形状

#sample from d-dimensional cube
def make_cubes(d, npoints, ncubes):
    cubes = []
    for i in range(ncubes):
        c = np.random.rand(npoints, d)     #sample from cube in d dimensions
        c += np.random.uniform(-5,5)   #random translation
        cubes.append(c)
       
    cubes = np.reshape(cubes, (npoints*ncubes,d) )
    return cubes

比如说

print( make_cubes(3,4,2) )

给我

[[-0.20605781  0.17759941 -0.54688093]
 [ 0.15476342 -0.47873276 -0.66528647]
 [ 0.05216092 -0.7118765  -0.2794354 ]
 [-0.648958   -0.49131152 -0.24827643]
 [ 3.80803203  3.41893782  3.66186498]
 [ 4.37535951  3.79385615  3.52051711]
 [ 3.98299149  4.17239746  4.09716118]
 [ 3.89030706  4.26959177  4.38166707]]