numpy：将矩阵行分配给另一个矩阵中的索引行的有效方法

Question

在以下代码中：

num_centroids = 8

vectors = # ndarray of shape (10000, 128)
centroids = np.random.uniform(-1., 1., (num_centroids   , vectors.shape[1]))

for _ in range(max_iterations):
    distances = sklearn.metrics.pairwise.euclidean_distances(vectors, centroids)

    # can this (the centroid update) be done any cleaner?
    for c in range(num_centroids):
        centroids[c] = np.mean(vectors[np.argmax(distances, axis=1) == c], axis=0)

是否有一种更有效或更干净的方法（除了for循环）将每个centroid（centroids矩阵中的一行）分配给从索引的{{1} }矩阵，假设vector索引（centroid）和使用c的向量索引是相同的？