GiulioRossetti / cdlib / 537

pending completion

Build # 537

Build Type

Pull #116

travis-ci

Committed by

web-flow

Commit Message

<a href="https://github.com/GiulioRossetti/cdlib/commit/<a class=hub.com/GiulioRossetti/cdlib/commit/<a class="double-link" href="https://git"><a class=hub.com/GiulioRossetti/cdlib/commit/<a class="double-link" href="https://git"><a class=hub.com/GiulioRossetti/cdlib/commit/<a class="double-link" href="https://git"><a class=hub.com/GiulioRossetti/cdlib/commit/d3c562e9ba24cdb62205546943e7368a8766c2fb">d3c562e9b">&lt;a href=&quot;https://github.com/GiulioRossetti/cdlib/commit/</a><a class="double-link" href="https://github.com/GiulioRossetti/cdlib/commit/&lt;a class=&quot;double-link&quot; href=&quot;https://git">&lt;a class=</a>hub.com/GiulioRossetti/cdlib/commit/&lt;a class=&quot;double-link&quot; href=&quot;https://git">&lt;a class=</a>hub.com/GiulioRossetti/cdlib/commit/&lt;a class=&quot;double-link&quot; href=&quot;https://git">&lt;a class=</a>hub.com/GiulioRossetti/cdlib/commit/d3c562e9ba24cdb62205546943e7368a8766c2fb">d3c562e9b</a><a href="https://github.com/GiulioRossetti/cdlib/commit/d3c562e9ba24cdb62205546943e7368a8766c2fb">&quot;&gt;&amp;lt;a href=&amp;quot;https://github.com/GiulioRossetti/cdlib/commit/&lt;/a&gt;&lt;a class=&quot;double-link&quot; href=&quot;https://github.com/GiulioRossetti/cdlib/commit/&amp;lt;a class=&amp;quot;double-link&amp;quot; href=&amp;quot;https://git&quot;&gt;&amp;lt;a class=&lt;/a&gt;hub.com/GiulioRossetti/cdlib/commit/&amp;lt;a class=&amp;quot;double-link&amp;quot; href=&amp;quot;https://git&quot;&gt;&amp;lt;a class=&lt;/a&gt;hub.com/GiulioRossetti/cdlib/commit/&amp;lt;a class=&amp;quot;double-link&amp;quot; href=&amp;quot;https://git&quot;&gt;&amp;lt;a class=&lt;/a&gt;hub.com/GiulioRossetti/cdlib/commit/&amp;lt;a class=&amp;quot;double-link&amp;quot; href=&amp;quot;https://git&quot;&gt;&amp;lt;a class=&lt;/a&gt;hub.com/GiulioRossetti/cdlib/commit/&amp;lt;a class=&amp;quot;double-link&amp;quot; href=&amp;quot;https://git&quot;&gt;&amp;lt;a class=&lt;/a&gt;hub.com/GiulioRossetti/cdlib/commit/d3c562e9ba24cdb62205546943e7368a8766c2fb&quot;&gt;d3c562e9b&lt;/a&gt;&lt;a href=&quot;https://github.com/GiulioRossetti/cdlib/commit/d3c562e9ba24cdb62205546943e7368a8766c2fb&quot;&gt;&amp;quot;&amp;gt;&amp;amp;lt;a href=&amp;amp;quot;https://github.com/GiulioRossetti/cdlib/commit/&amp;lt;/a&amp;gt;&amp;lt;a class=&amp;quot;double-link&a... (continued)

Pull Request Pull Request #116: fixed plot_scoring

Run Details

19 of 19 new or added lines in 2 files covered. (100.0%)

2188 of 2330 relevant lines covered (93.91%)

1.88 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

76.67

/cdlib/viz/plots.py

import seaborn as sns
import pandas as pd
import cdlib
import matplotlib.pyplot as plt

__all__ = ["plot_com_properties_relation", "plot_com_stat", "plot_sim_matrix", "plot_scoring"]


def plot_sim_matrix(clusterings, scoring):
    """
    Plot a similarity matrix between a list of clusterings, using the provided scoring function.

    :param clusterings: list of clusterings to compare
    :param scoring: the scoring function to use
    :return: a ClusterGrid instance

    Example:

    >>> from cdlib import algorithms, viz, evaluation
    >>> import networkx as nx
    >>> g = nx.karate_club_graph()
    >>> coms = algorithms.louvain(g)
    >>> coms2 = algorithms.walktrap(g)
    >>> clustermap = viz.plot_sim_matrix([coms,coms2],evaluation.adjusted_mutual_information)

    """
    forDF = []
    for c in clusterings:
        cID = c.get_description()
        for c2 in clusterings:
            c2ID = c2.get_description()
            forDF.append([cID, c2ID, scoring(c, c2).score])
    df = pd.DataFrame(columns=["com1", "com2", "score"], data=forDF)
    df = df.pivot("com1", "com2", "score")
    return sns.clustermap(df)


def plot_com_stat(com_clusters, com_fitness):
    """
    Plot the distribution of a property among all communities for a clustering, or a list of clusterings (violin-plots)

    :param com_clusters: list of clusterings to compare, or a single clustering
    :param com_fitness: the fitness/community property to use
    :return: the violin-plots

    Example:

    >>> from cdlib import algorithms, viz, evaluation
    >>> import networkx as nx
    >>> g = nx.karate_club_graph()
    >>> coms = algorithms.louvain(g)
    >>> coms2 = algorithms.walktrap(g)
    >>> violinplot = viz.plot_com_stat([coms,coms2],evaluation.size)

    """
    if isinstance(com_clusters, cdlib.classes.clustering.Clustering):
        com_clusters = [com_clusters]

    allVals = []
    allNames = []
    for c in com_clusters:
        prop = com_fitness(c.graph, c, summary=False)
        allVals += prop
        allNames += [c.get_description()] * len(prop)

    ax = sns.violinplot(allNames, allVals,cut=0,saturation=0.5,palette="Set3")
    for tick in ax.get_xticklabels():
        tick.set_rotation(90)

    plt.ylabel("%s" % com_fitness.__name__)
    plt.xlabel("Algorithm")
    plt.tight_layout()




    return ax


def plot_com_properties_relation(com_clusters, com_fitness_x, com_fitness_y, **kwargs):
    """
    Plot the relation between two properties/fitness function of a clustering

    :param com_clusters:  clustering(s) to analyze (cluster or cluster list)
    :param com_fitness_x: first fitness/community property
    :param com_fitness_y: first fitness/community property
    :param kwargs: parameters for the seaborn lmplot
    :return: a seaborn lmplot

    Example:

    >>> from cdlib import algorithms, viz, evaluation
    >>> import networkx as nx
    >>> g = nx.karate_club_graph()
    >>> coms = algorithms.louvain(g)
    >>> coms2 = algorithms.walktrap(g)
    >>> lmplot = viz.plot_com_properties_relation([coms,coms2],evaluation.size,evaluation.internal_edge_density)
    """
    if isinstance(com_clusters, cdlib.classes.clustering.Clustering):
        com_clusters = [com_clusters]

    for_df = []

    for c in com_clusters:
        x = com_fitness_x(c.graph, c, summary=False)
        y = com_fitness_y(c.graph, c, summary=False)
        for i, vx in enumerate(x):
            for_df.append([c.get_description(), vx, y[i]])

    df = pd.DataFrame(columns=["Method", com_fitness_x.__name__, com_fitness_y.__name__], data=for_df)
    ax = sns.lmplot(x=com_fitness_x.__name__, y=com_fitness_y.__name__, data=df, hue="Method", fit_reg=False,legend=False, x_bins=100,**kwargs)
    plt.legend(loc='best')

    # if log_x:
    #     ax.set_xscale("log")
    # if log_y:
    #     ax.set_yscale("log")
    plt.tight_layout()

    return ax


def plot_scoring(graphs, ref_partitions, graph_names, methods, scoring=cdlib.evaluation.adjusted_mutual_information,
                nbRuns=5):
   """
   Plot the scores obtained by a list of methods on a list of graphs.

   :param graphs: list of graphs on which to make computations
   :param ref_partitions: list of reference clusterings corresponding to graphs
   :param graph_names: list of the names of the graphs to display
   :param methods: list of functions that take a graph as input and return a Clustering as output
   :param scoring: the scoring function to use, default anmi
   :param nbRuns: number of runs to do for each method on each graph
   :return: a seaborn lineplot

   Example:

   >>> from cdlib import algorithms, viz, evaluation
   >>> import networkx as nx
   >>> g1 = nx.algorithms.community.LFR_benchmark_graph(1000, 3, 1.5, 0.5, min_community=20, average_degree=5)
   >>> g2 = nx.algorithms.community.LFR_benchmark_graph(1000, 3, 1.5, 0.7, min_community=20, average_degree=5)
   >>> names = ["g1", "g2"]
   >>> graphs = [g1, g2]
   >>> for g in graphs:
   >>>     references.append(NodeClustering(communities={frozenset(g.nodes[v]['community']) for v in g}, graph=g, method_name="reference"))
   >>> algos = [algorithms.crisp_partition.louvain, algorithms.crisp_partition.label_propagation]
   >>> viz.plot_scoring(graphs, references, names, algos, nbRuns=2)

   """
   forDF = []
   for i, g in enumerate(graphs):
       for m in methods:
           for r in range(nbRuns):
               partition = m(g)

               score = scoring(partition, ref_partitions[i]).score
               forDF.append([graph_names[i], score, partition.get_description()])
   df = pd.DataFrame(columns=["graph", "score", "method"], data=forDF)
   ax = sns.lineplot(x="graph", y="score", hue="method", data=df, legend="brief")
   ax.legend(loc='best')
   for tick in ax.get_xticklabels():
       tick.set_rotation(90)
   plt.tight_layout()

   return ax

1	import seaborn as sns	2✔
2	import pandas as pd	2✔
3	import cdlib	2✔
4	import matplotlib.pyplot as plt	2✔
5
6	__all__ = ["plot_com_properties_relation", "plot_com_stat", "plot_sim_matrix", "plot_scoring"]	2✔
7
8
9	def plot_sim_matrix(clusterings, scoring):	2✔
10	"""
11	Plot a similarity matrix between a list of clusterings, using the provided scoring function.
12
13	:param clusterings: list of clusterings to compare
14	:param scoring: the scoring function to use
15	:return: a ClusterGrid instance
16
17	Example:
18
19	>>> from cdlib import algorithms, viz, evaluation
20	>>> import networkx as nx
21	>>> g = nx.karate_club_graph()
22	>>> coms = algorithms.louvain(g)
23	>>> coms2 = algorithms.walktrap(g)
24	>>> clustermap = viz.plot_sim_matrix([coms,coms2],evaluation.adjusted_mutual_information)
25
26	"""
27	forDF = []	2✔
28	for c in clusterings:	2✔
29	cID = c.get_description()	2✔
30	for c2 in clusterings:	2✔
31	c2ID = c2.get_description()	2✔
32	forDF.append([cID, c2ID, scoring(c, c2).score])	2✔
33	df = pd.DataFrame(columns=["com1", "com2", "score"], data=forDF)	2✔
34	df = df.pivot("com1", "com2", "score")	2✔
35	return sns.clustermap(df)	2✔
36
37
38	def plot_com_stat(com_clusters, com_fitness):	2✔
39	"""
40	Plot the distribution of a property among all communities for a clustering, or a list of clusterings (violin-plots)
41
42	:param com_clusters: list of clusterings to compare, or a single clustering
43	:param com_fitness: the fitness/community property to use
44	:return: the violin-plots
45
46	Example:
47
48	>>> from cdlib import algorithms, viz, evaluation
49	>>> import networkx as nx
50	>>> g = nx.karate_club_graph()
51	>>> coms = algorithms.louvain(g)
52	>>> coms2 = algorithms.walktrap(g)
53	>>> violinplot = viz.plot_com_stat([coms,coms2],evaluation.size)
54
55	"""
56	if isinstance(com_clusters, cdlib.classes.clustering.Clustering):	2✔
57	com_clusters = [com_clusters]	2✔
58
59	allVals = []	2✔
60	allNames = []	2✔
61	for c in com_clusters:	2✔
62	prop = com_fitness(c.graph, c, summary=False)	2✔
63	allVals += prop	2✔
64	allNames += [c.get_description()] * len(prop)	2✔
65
66	ax = sns.violinplot(allNames, allVals,cut=0,saturation=0.5,palette="Set3")	2✔
67	for tick in ax.get_xticklabels():	2✔
68	tick.set_rotation(90)	2✔
69
70	plt.ylabel("%s" % com_fitness.__name__)	2✔
71	plt.xlabel("Algorithm")	2✔
72	plt.tight_layout()	2✔
73
74
75
76
77	return ax	2✔
78
79
80	def plot_com_properties_relation(com_clusters, com_fitness_x, com_fitness_y, **kwargs):	2✔
81	"""
82	Plot the relation between two properties/fitness function of a clustering
83
84	:param com_clusters: clustering(s) to analyze (cluster or cluster list)
85	:param com_fitness_x: first fitness/community property
86	:param com_fitness_y: first fitness/community property
87	:param kwargs: parameters for the seaborn lmplot
88	:return: a seaborn lmplot
89
90	Example:
91
92	>>> from cdlib import algorithms, viz, evaluation
93	>>> import networkx as nx
94	>>> g = nx.karate_club_graph()
95	>>> coms = algorithms.louvain(g)
96	>>> coms2 = algorithms.walktrap(g)
97	>>> lmplot = viz.plot_com_properties_relation([coms,coms2],evaluation.size,evaluation.internal_edge_density)
98	"""
99	if isinstance(com_clusters, cdlib.classes.clustering.Clustering):	2✔
100	com_clusters = [com_clusters]	2✔
101
102	for_df = []	2✔
103
104	for c in com_clusters:	2✔
105	x = com_fitness_x(c.graph, c, summary=False)	2✔
106	y = com_fitness_y(c.graph, c, summary=False)	2✔
107	for i, vx in enumerate(x):	2✔
108	for_df.append([c.get_description(), vx, y[i]])	2✔
109
110	df = pd.DataFrame(columns=["Method", com_fitness_x.__name__, com_fitness_y.__name__], data=for_df)	2✔
111	ax = sns.lmplot(x=com_fitness_x.__name__, y=com_fitness_y.__name__, data=df, hue="Method", fit_reg=False,legend=False, x_bins=100,**kwargs)	2✔
112	plt.legend(loc='best')	2✔
113
114	# if log_x:
115	# ax.set_xscale("log")
116	# if log_y:
117	# ax.set_yscale("log")
118	plt.tight_layout()	2✔
119
120	return ax	2✔
121
122
123	def plot_scoring(graphs, ref_partitions, graph_names, methods, scoring=cdlib.evaluation.adjusted_mutual_information,	2✔
124	nbRuns=5):
125	"""
126	Plot the scores obtained by a list of methods on a list of graphs.
127
128	:param graphs: list of graphs on which to make computations
129	:param ref_partitions: list of reference clusterings corresponding to graphs
130	:param graph_names: list of the names of the graphs to display
131	:param methods: list of functions that take a graph as input and return a Clustering as output
132	:param scoring: the scoring function to use, default anmi
133	:param nbRuns: number of runs to do for each method on each graph
134	:return: a seaborn lineplot
135
136	Example:
137
138	>>> from cdlib import algorithms, viz, evaluation
139	>>> import networkx as nx
140	>>> g1 = nx.algorithms.community.LFR_benchmark_graph(1000, 3, 1.5, 0.5, min_community=20, average_degree=5)
141	>>> g2 = nx.algorithms.community.LFR_benchmark_graph(1000, 3, 1.5, 0.7, min_community=20, average_degree=5)
142	>>> names = ["g1", "g2"]
143	>>> graphs = [g1, g2]
144	>>> for g in graphs:
145	>>> references.append(NodeClustering(communities={frozenset(g.nodes[v]['community']) for v in g}, graph=g, method_name="reference"))
146	>>> algos = [algorithms.crisp_partition.louvain, algorithms.crisp_partition.label_propagation]
147	>>> viz.plot_scoring(graphs, references, names, algos, nbRuns=2)
148
149	"""
150	forDF = []	×
151	for i, g in enumerate(graphs):	×
152	for m in methods:	×
153	for r in range(nbRuns):	×
154	partition = m(g)	×
155
156	score = scoring(partition, ref_partitions[i]).score	×
157	forDF.append([graph_names[i], score, partition.get_description()])	×
158	df = pd.DataFrame(columns=["graph", "score", "method"], data=forDF)	×
159	ax = sns.lineplot(x="graph", y="score", hue="method", data=df, legend="brief")	×
160	ax.legend(loc='best')	×
161	for tick in ax.get_xticklabels():	×
162	tick.set_rotation(90)	×
163	plt.tight_layout()	×
164
165	return ax	×

GiulioRossetti / cdlib / 537

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous