def compute_metrics(G, refresh=True):
descriptTLU = 'TLU--Local Clustering Coefficient'
descriptDegree = 'Degree Statistics'
if refresh or G.metrics == None:
#we treat a single node graph to have a density of 1
#TODO: This is undefined for multigraphs. Prob should simplify if this happens
density = G.density() if G.vcount() > 1 else 1.0
G.metrics = {
'Internal Number Nodes':
G.vcount(),
'Internal Number Edges':
G.ecount(),
'Density':
density,
'Diameter':
G.diameter(),
'Cohesiveness':
G.cohesiveness(),
'Triangle Participation Ratio':
G.triangle_participation_ratio(),
'Transitivity Undirected (Global Clustering Coefficient)':
G.transitivity_undirected(mode='zero')
}
G.metrics.update(
aggregate(G.transitivity_local_undirected(mode='zero'),
prefix=descriptTLU))
G.metrics.update(aggregate(G.degree(), prefix=descriptDegree))
def compute_metrics(G, refresh = True):
descriptTLU = 'TLU--Local Clustering Coefficient'
descriptDegree = 'Degree Statistics'
if refresh or G.metrics == None:
G.metrics = {
'Internal Number Nodes' : G.vcount(),
'Internal Number Edges' : G.ecount(),
'Density' : G.density(),
'Diameter' : G.diameter(),
'Cohesiveness' : G.cohesiveness(),
'Triangle Participation Ratio' : G.triangle_participation_ratio(),
'Transitivity Undirected (Global Clustering Coefficient)'
: G.transitivity_undirected()
}
G.metrics.update(aggregate(G.transitivity_local_undirected(mode='zero'), prefix=descriptTLU))
G.metrics.update(aggregate(G.degree(), prefix=descriptDegree))
def compute_metrics(G, refresh = True):
descriptTLU = 'TLU--Local Clustering Coefficient'
descriptDegree = 'Degree Statistics'
if refresh or G.metrics == None:
#we treat a single node graph to have a density of 1
#TODO: This is undefined for multigraphs. Prob should simplify if this happens
density = G.density() if G.vcount() > 1 else 1.0
G.metrics = {
'Internal Number Nodes' : G.vcount(),
'Internal Number Edges' : G.ecount(),
'Density' : density,
'Diameter' : G.diameter(),
'Cohesiveness' : G.cohesiveness(),
'Triangle Participation Ratio' : G.triangle_participation_ratio(),
'Transitivity Undirected (Global Clustering Coefficient)'
: G.transitivity_undirected(mode='zero')
}
G.metrics.update(aggregate(G.transitivity_local_undirected(mode='zero'), prefix=descriptTLU))
G.metrics.update(aggregate(G.degree(), prefix=descriptDegree))
def compute_metrics(cover, weights=None, ground_truth_cover=None):
t0 = time.time()
fomd_results = fomd(cover, weights)
expansion_results = expansion(cover, weights)
cut_ratio_results = cut_ratio(cover)
conductance_results = conductance(cover, weights)
n_cut_results = normalized_cut(cover, weights)
max_out_results = maximum_out_degree_fraction(cover, weights)
avg_out_results = average_out_degree_fraction(cover, weights)
flake_odf_results = flake_out_degree_fraction(cover, weights)
sep_results = separability(cover,weights)
results_key = "results"
agg_key = "aggegations"
cover.metrics = {
'Fraction over a Median Degree' : {results_key:fomd_results, agg_key:aggregate(fomd_results)},
'Expansion' : {results_key:expansion_results, agg_key:aggregate(expansion_results)},
'Cut Ratio' : {results_key:cut_ratio_results, agg_key:aggregate(cut_ratio_results)},
'Conductance' : {results_key:conductance_results, agg_key:aggregate(conductance_results)},
'Normalized Cut' : {results_key:n_cut_results, agg_key:aggregate(n_cut_results)},
'Maximum Out Degree Fraction' : {results_key:max_out_results, agg_key:aggregate(max_out_results)},
'Average Out Degree Fraction' : {results_key:avg_out_results, agg_key:aggregate(avg_out_results)},
'Flake Out Degree Fraction' : {results_key:flake_odf_results, agg_key:aggregate(flake_odf_results)},
'Separability' : {results_key:sep_results, agg_key:aggregate(sep_results)},
}
for i in range(len(cover)):
sg = cover.subgraph(i)
sg.compute_metrics(refresh=False)
#we want to add the metrics from the subgraph calculations to the current cover. The cover and
#subgraph are essentially the same thing, however because we use the igraph graph functions we
#can't natively call these call a cover... hence the need to transfer over the results
for key, val in sg.metrics.items():
if key not in cover.metrics:
cover.metrics[key] = {results_key:[], agg_key:None}
cover.metrics[key][results_key] += [val]
#aggregate just the results from the subgraph metrics
for k in sg.metrics.keys():
cover.metrics[k][agg_key] = aggregate(cover.metrics[k][results_key])
cover.metrics['omega'] = compare_omega(cover, ground_truth_cover)
cover.metrics['metrics_total_time'] = time.time() - t0
def compute_metrics(cover, weights=None, ground_truth_cover=None):
t0 = time.time()
fomd_results = fomd(cover, weights)
expansion_results = expansion(cover, weights)
cut_ratio_results = cut_ratio(cover)
conductance_results = conductance(cover, weights)
n_cut_results = normalized_cut(cover, weights)
# Get out degree fraction results then reuse for max, average, flake
odf = out_degree_fraction(cover, weights)
odf = odf.tocsc()
max_out_results = maximum_out_degree_fraction(cover,
odf=odf,
weights=weights)
avg_out_results = average_out_degree_fraction(cover,
odf=odf,
weights=weights)
flake_odf_results = flake_out_degree_fraction(cover,
odf=odf,
weights=weights)
sep_results = separability(cover, weights)
results_key = "results"
agg_key = "aggregations"
cover.metrics = {
'Fraction over a Median Degree': {
results_key: fomd_results,
agg_key: aggregate(fomd_results)
},
'Expansion': {
results_key: expansion_results,
agg_key: aggregate(expansion_results)
},
'Cut Ratio': {
results_key: cut_ratio_results,
agg_key: aggregate(cut_ratio_results)
},
'Conductance': {
results_key: conductance_results,
agg_key: aggregate(conductance_results)
},
'Normalized Cut': {
results_key: n_cut_results,
agg_key: aggregate(n_cut_results)
},
'Maximum Out Degree Fraction': {
results_key: max_out_results,
agg_key: aggregate(max_out_results)
},
'Average Out Degree Fraction': {
results_key: avg_out_results,
agg_key: aggregate(avg_out_results)
},
'Flake Out Degree Fraction': {
results_key: flake_odf_results,
agg_key: aggregate(flake_odf_results)
},
'Separability': {
results_key: sep_results,
agg_key: aggregate(sep_results)
},
}
for i in range(len(cover)):
sg = cover.subgraph(i)
sg.compute_metrics(refresh=False)
#we want to add the metrics from the subgraph calculations to the current cover. The cover and
#subgraph are essentially the same thing, however because we use the igraph graph functions we
#can't natively call these call a cover... hence the need to transfer over the results
for key, val in sg.metrics.items():
if key not in cover.metrics:
cover.metrics[key] = {results_key: [], agg_key: None}
cover.metrics[key][results_key] += [val]
#aggregate just the results from the subgraph metrics
for k in sg.metrics.keys():
cover.metrics[k][agg_key] = aggregate(cover.metrics[k][results_key])
