def test_adjust_partition():
e = np.loadtxt(os.path.join(os.path.dirname(__file__), 'jazz.net'),
skiprows=3, dtype=int)[:, :2] - 1
g = nx.Graph()
g.add_edges_from(e)
p0 = mod.newman_partition(g)
def test_adjust_partition():
e = np.loadtxt(os.path.join(os.path.dirname(__file__), 'jazz.net'),
skiprows=3,
dtype=int)[:, :2] - 1
g = nx.Graph()
g.add_edges_from(e)
p0 = mod.newman_partition(g)
def graph_analysis (graph):
part = modularity.newman_partition(graph)
print part.modularity()
part = util.dictset_to_listset(part.index)
part = weighted_modularity.WeightedPartition(graph,part)
pc = np.array(participation_coefficient(part,catch_edgeless_node=False).values())
wmd = np.array(within_community_degree(part,catch_edgeless_node=False).values())
return pc,wmd
def test_adjust_partition():
e = np.loadtxt(os.path.join(os.path.dirname(__file__), 'jazz.net'),
skiprows=3, dtype=int)[:, :2] - 1
g = nx.Graph()
g.add_edges_from(e)
p0 = mod.newman_partition(g)
p1 = mod.adjust_partition(g, p0, max_iter=6)
npt.assert_(p0 > 0.38)
npt.assert_(p1 > 0.42)
def test_adjust_partition():
e = np.loadtxt(os.path.join(os.path.dirname(__file__), 'jazz.net'),
skiprows=3,
dtype=int)[:, :2] - 1
g = nx.Graph()
g.add_edges_from(e)
p0 = mod.newman_partition(g)
p1 = mod.adjust_partition(g, p0, max_iter=6)
npt.assert_(p0 > 0.38)
npt.assert_(p1 > 0.42)
def test_newman_partition():
""" Test Newman Partition function """
tmpmat = np.random.random((10, 10))
tmpmat[tmpmat < .5] = 0
graph = nx.from_numpy_matrix(tmpmat, nx.Graph(weighted=False))
npt.assert_raises(ValueError, mod.newman_partition, graph)
tmpmat[:] = 0
# test that no edges raises error (from GraphPartition)
graph = nx.from_numpy_matrix(tmpmat, nx.Graph(weighted=False))
npt.assert_raises(ValueError, mod.newman_partition, graph)
tmpmat[:] = 1
util.fill_diagonal(tmpmat, 0)
graph = nx.from_numpy_matrix(tmpmat, nx.Graph(weighted=False))
part = mod.newman_partition(graph)
## if all edges are connected expect only one partition
expected_part = {0: set([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])}
nt.assert_equal(part.index, expected_part)
def test_newman_partition():
""" Test Newman Partition function """
tmpmat = np.random.random((10, 10))
tmpmat[tmpmat < 0.5] = 0
graph = nx.from_numpy_matrix(tmpmat, nx.Graph(weighted=False))
npt.assert_raises(ValueError, mod.newman_partition, graph)
tmpmat[:] = 0
# test that no edges raises error (from GraphPartition)
graph = nx.from_numpy_matrix(tmpmat, nx.Graph(weighted=False))
npt.assert_raises(ValueError, mod.newman_partition, graph)
tmpmat[:] = 1
util.fill_diagonal(tmpmat, 0)
graph = nx.from_numpy_matrix(tmpmat, nx.Graph(weighted=False))
part = mod.newman_partition(graph)
## if all edges are connected expect only one partition
expected_part = {0: set([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])}
nt.assert_equal(part.index, expected_part)
def main(argv=sys.argv):
filename = argv[
1] # file name convention: cmat_<project>_<atlas>_<subgroup>_S<subject#>_B<block#>_<cost>.npy
failedcmd = argv[2] # newman or adjust
# Load saved cmat generated from NewmanMod.py
cmat_1 = np.load(filename)
# newman.partition fails if there are Inf/nan's in cmat_1
# this occurs on the diagonal (Z-transformed autocorrelation == Inf)
# so we'll convert the diagonal to 0
# NOTE: this manipulation to cmat_1 also changes modularity values
# calculated with simulated annealing - we don't understand why this happens
temp = np.isnan(cmat_1)
cmat_1[temp] = 0
G1 = nx.Graph(weighted=False)
G1 = nx.from_numpy_matrix(cmat_1, G1)
P1 = md.newman_partition(G1)
if failedcmd == 'adjust':
P1 = md.adjust_partition(G1, P1)
print('Modularity: ', P1.modularity())
#save partition information (nodelist is not relabeled here)
pname = 'part_norelabel_' + '_'.join(filename.split('_')[1:])
pout = open(pname, 'w')
pickle.dump(P1.index, pout)
pout.close()
print('Wrote ', pname)
#clear variables
del P1, cmat_1, G1
def calc_spectral_modularity(mat, cost):
""" Use brainx to
generate binary matrix theshold based on cost
run newman spectral modularity
Parameters
----------
mat : numpy matrix
symmetric adjacency matrix, should not contain nan
cost: float
cost used to binarize adjacency matrix
Returns
-------
part : GraphPartition
part.index provides access to found communities
true_cost : float
the actual cost associated with thresholded adj matrix
"""
mask, real_cost = util.threshold_adjacency_matrix(mat, cost)
true_cost = util.find_true_cost(mask)
graph = nx.from_numpy_matrix(mask)
part = modularity.newman_partition(graph)
return part, true_cost
def main(argv = sys.argv):
filename = argv[1] # file name convention: cmat_<project>_<atlas>_<subgroup>_S<subject#>_B<block#>_<cost>.npy
failedcmd = argv[2] # newman or adjust
# Load saved cmat generated from NewmanMod.py
cmat_1 = np.load(filename)
# newman.partition fails if there are Inf/nan's in cmat_1
# this occurs on the diagonal (Z-transformed autocorrelation == Inf)
# so we'll convert the diagonal to 0
# NOTE: this manipulation to cmat_1 also changes modularity values
# calculated with simulated annealing - we don't understand why this happens
temp=np.isnan(cmat_1)
cmat_1[temp]=0
G1 = nx.Graph(weighted = False)
G1 = nx.from_numpy_matrix(cmat_1,G1)
P1 = md.newman_partition(G1)
if failedcmd == 'adjust':
P1 = md.adjust_partition(G1,P1)
print('Modularity: ', P1.modularity())
#save partition information (nodelist is not relabeled here)
pname = 'part_norelabel_' + '_'.join(filename.split('_')[1:])
pout = open(pname,'w')
pickle.dump(P1.index,pout)
pout.close()
print('Wrote ',pname)
#clear variables
del P1, cmat_1, G1
#!/usr/bin/env python
"""This code is crashing with:
IndexError: index 32 is out of bounds for axis 1 with size 25
"""
import cPickle as pickle
from brainx import modularity as md
G3 = pickle.load(open("G3.pck", "rb"))
P3 = md.newman_partition(G3)
