Louvain 层次结构

此笔记本说明了通过 Louvain（自下而上的连续聚合）对图进行层次聚类。

[1]:

from IPython.display import SVG

[2]:

import numpy as np

[3]:

from sknetwork.data import karate_club, painters, movie_actor
from sknetwork.hierarchy import LouvainHierarchy
from sknetwork.hierarchy import cut_straight, dasgupta_score, tree_sampling_divergence
from sknetwork.visualization import visualize_graph, visualize_bigraph, visualize_dendrogram

图

[4]:

graph = karate_club(metadata=True)
adjacency = graph.adjacency
position = graph.position

[5]:

# hierarchical clustering
louvain = LouvainHierarchy()
dendrogram = louvain.fit_predict(adjacency)

[6]:

image = visualize_dendrogram(dendrogram)
SVG(image)

[6]:

../../_images/tutorials_hierarchy_louvain_recursion_8_0.svg

[7]:

# cuts
labels = cut_straight(dendrogram)
print(labels)

[0 0 0 0 3 3 3 0 1 0 3 0 0 0 1 1 3 0 1 0 1 0 1 2 2 2 1 2 2 1 1 2 1 1]

[8]:

labels, dendrogram_aggregate = cut_straight(dendrogram, n_clusters=4, return_dendrogram=True)
print(labels)

[0 0 0 0 3 3 3 0 1 0 3 0 0 0 1 1 3 0 1 0 1 0 1 2 2 2 1 2 2 1 1 2 1 1]

[9]:

_, counts = np.unique(labels, return_counts=True)

[10]:

image = visualize_dendrogram(dendrogram_aggregate, names=counts, rotate_names=False)
SVG(image)

[10]:

../../_images/tutorials_hierarchy_louvain_recursion_12_0.svg

[11]:

image = visualize_graph(adjacency, position, labels=labels)
SVG(image)

[11]:

../../_images/tutorials_hierarchy_louvain_recursion_13_0.svg

[12]:

# metrics
dasgupta_score(adjacency, dendrogram)

[12]:

0.6496983408748114

有向图

[13]:

graph = painters(metadata=True)
adjacency = graph.adjacency
position = graph.position
names = graph.names

[14]:

# hierarchical clustering
louvain = LouvainHierarchy()
dendrogram = louvain.fit_predict(adjacency)

[15]:

image = visualize_dendrogram(dendrogram, names, rotate=True)
SVG(image)

[15]:

../../_images/tutorials_hierarchy_louvain_recursion_18_0.svg

[16]:

# cut with 3 clusters
labels = cut_straight(dendrogram, n_clusters = 3)
print(labels)

[0 1 2 1 2 2 0 1 0 2 0 1 1 0]

[17]:

image = visualize_graph(adjacency, position, names=names, labels=labels)
SVG(image)

[17]:

../../_images/tutorials_hierarchy_louvain_recursion_20_0.svg

[18]:

# metrics
dasgupta_score(adjacency, dendrogram)

[18]:

0.52

二部图

[19]:

graph = movie_actor(metadata=True)
biadjacency = graph.biadjacency
names_row = graph.names_row
names_col = graph.names_col

[20]:

# hierarchical clustering
louvain = LouvainHierarchy()
louvain.fit(biadjacency)
dendrogram_row = louvain.dendrogram_row_
dendrogram_col = louvain.dendrogram_col_
dendrogram_full = louvain.dendrogram_full_

[21]:

image = visualize_dendrogram(dendrogram_row, names_row, n_clusters=4, rotate=True)
SVG(image)

[21]:

../../_images/tutorials_hierarchy_louvain_recursion_25_0.svg

[22]:

image = visualize_dendrogram(dendrogram_col, names_col, n_clusters=4, rotate=True)
SVG(image)

[22]:

../../_images/tutorials_hierarchy_louvain_recursion_26_0.svg

[23]:

# cuts
labels = cut_straight(dendrogram_full, n_clusters = 4)
n_row = biadjacency.shape[0]
labels_row = labels[:n_row]
labels_col = labels[n_row:]

[24]:

image = visualize_bigraph(biadjacency, names_row, names_col, labels_row, labels_col)
SVG(image)

[24]:

../../_images/tutorials_hierarchy_louvain_recursion_28_0.svg