TLDR:重要的是,问题是在二维中设置的。对于大尺寸,这些技术可能无效。
在 2D 中,我们可以在 O(n log n) 时间计算每个簇的直径(簇内距离),其中 n 是使用凸包的簇大小。矢量化用于加速剩余操作。文末提到了两种可能的渐近改进,欢迎投稿;)
设置和伪造数据:
import numpy as np
from scipy import spatial
from matplotlib import pyplot as plt
# set up fake data
np.random.seed(0)
n_centroids = 1000
centroids = np.random.rand(n_centroids, 2)
cluster_sizes = np.random.randint(1, 1000, size=n_centroids)
# labels from 1 to n_centroids inclusive
labels = np.repeat(np.arange(n_centroids), cluster_sizes) + 1
points = np.zeros((cluster_sizes.sum(), 2))
points[:,0] = np.repeat(centroids[:,0], cluster_sizes)
points[:,1] = np.repeat(centroids[:,1], cluster_sizes)
points += 0.05 * np.random.randn(cluster_sizes.sum(), 2)
看起来有点像这样:
接下来,我们定义了一个 diameter 函数,用于计算最大的簇内距离,基于使用凸包的 this 方法。
# compute the diameter based on convex hull
def diameter(pts):
# need at least 3 points to construct the convex hull
if pts.shape[0] <= 1:
return 0
if pts.shape[0] == 2:
return ((pts[0] - pts[1])**2).sum()
# two points which are fruthest apart will occur as vertices of the convex hull
hull = spatial.ConvexHull(pts)
candidates = pts[spatial.ConvexHull(pts).vertices]
return spatial.distance_matrix(candidates, candidates).max()
对于 Dunn 指数计算,我假设我们已经计算了点、聚类标签和聚类质心。
如果集群的数量很大,以下基于 Pandas 的解决方案可能会表现良好:
import pandas as pd
def dunn_index_pandas(pts, labels, centroids):
# O(k n log(n)) with k clusters and n points; better performance with more even clusters
max_intracluster_dist = pd.DataFrame(pts).groupby(labels).agg(diameter_pandas)[0].max()
# O(k^2) with k clusters; can be reduced to O(k log(k))
# get pairwise distances between centroids
cluster_dmat = spatial.distance_matrix(centroids, centroids)
# fill diagonal with +inf: ignore zero distance to self in "min" computation
np.fill_diagonal(cluster_dmat, np.inf)
min_intercluster_dist = cluster_sizes.min()
return min_intercluster_dist / max_intracluster_dist
否则,我们可以继续使用纯 numpy 解决方案。
def dunn_index(pts, labels, centroids):
# O(k n log(n)) with k clusters and n points; better performance with more even clusters
max_intracluster_dist = max(diameter(pts[labels==i]) for i in np.unique(labels))
# O(k^2) with k clusters; can be reduced to O(k log(k))
# get pairwise distances between centroids
cluster_dmat = spatial.distance_matrix(centroids, centroids)
# fill diagonal with +inf: ignore zero distance to self in "min" computation
np.fill_diagonal(cluster_dmat, np.inf)
min_intercluster_dist = cluster_sizes.min()
return min_intercluster_dist / max_intracluster_dist
%time dunn_index(points, labels, centroids)
# returned value 2.15
# in 2.2 seconds
%time dunn_index_pandas(points, labels, centroids)
# returned 2.15
# in 885 ms
对于具有 i.i.d. ~U[1,1000] 集群大小的 1000 集群,这需要 2.2。在我的机器上秒。对于此示例(许多小集群),使用 Pandas 方法,这个数字下降到 0.8 秒。
当集群数量很大时,还有两个相关的进一步优化机会:
首先,我使用蛮力 O(k^2) 方法计算最小集群间距离,其中 k 是集群的数量。如here 所讨论的,这可以简化为O(k log(k))。
其次,max(diameter(pts[labels==i]) for i in np.unique(labels)) 需要 k 传递大小为 n 的数组。对于许多集群,这可能成为瓶颈(如本例所示)。使用 pandas 方法可以稍微缓解这种情况,但我希望这可以进一步优化。对于当前参数,大约三分之一的计算时间花费在计算簇内距离的簇间之外。