gat*_*tor 4 artificial-intelligence cluster-analysis k-means python-3.x
邓恩指数是一种评估聚类的方法。值越高越好。它的计算方法是用最低簇间距离(即任何两个簇质心之间的最小距离)除以最高簇内距离(即任何簇中任意两点之间的最大距离)。
我有一个用于计算邓恩指数的代码片段:
def dunn_index(pf, cf):
"""
pf -- all data points
cf -- cluster centroids
"""
numerator = inf
for c in cf: # for each cluster
for t in cf: # for each cluster
if t is c: continue # if same cluster, ignore
numerator = min(numerator, distance(t, c)) # find distance between centroids
denominator = 0
for c in cf: # for each cluster
for p in pf: # for each point
if p.get_cluster() is not c: continue # if point not in cluster, ignore
for t in pf: # for each point
if t.get_cluster() is not c: continue # if point not in cluster, ignore
if t is p: continue # if same point, ignore
denominator = max(denominator, distance(t, p))
return numerator/denominator
Run Code Online (Sandbox Code Playgroud)
问题是这非常慢:对于由 5000 个实例和 15 个集群组成的示例数据集,上面的函数最坏需要执行超过 3.75 亿次距离计算。实际上它要低得多,但即使是数据已经按集群排序的最佳情况,也需要大约 2500 万次距离计算。我想减少它的时间,我已经尝试过直线距离与欧几里得,但效果不佳。
我怎样才能改进这个算法?
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,我们可以及时计算每个集群的直径(集群内距离),O(n log n)其中n使用凸包的集群大小。矢量化用于加速剩余操作。文章末尾提到了两种可能的渐进改进,欢迎贡献;)
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)
Run Code Online (Sandbox Code Playgroud)
看起来有点像这样:
接下来,我们diameter基于这种使用凸包的方法定义了一个用于计算最大簇内距离的函数。
# 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()
Run Code Online (Sandbox Code Playgroud)
对于邓恩指数计算,我假设我们已经计算了点、聚类标签和聚类质心。
如果集群数量较多,以下基于 Pandas 的方案可能表现良好:
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
Run Code Online (Sandbox Code Playgroud)
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
Run Code Online (Sandbox Code Playgroud)
对于1000具有i.i.d. ~U[1,1000]集群大小的集群,这需要 2.2。秒在我的机器上。对于本示例(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,还有两个相关的优化机会:
首先,我正在使用蛮力O(k^2)方法计算最小k集群间距离,其中是集群数量。这可以简化为O(k log(k)),正如这里所讨论的。
其次,max(diameter(pts[labels==i]) for i in np.unique(labels))需要k传递一个大小为 的数组n。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在计算集群内距离的集群之外。
| 归档时间: |
|
| 查看次数: |
2269 次 |
| 最近记录: |