def initial_dist_sim_pairs_with_netcomp(self, gs1, gs2):
if logging_enabled == True:
print(
"- Entered dist_sim_calculator::DistSimCalculator::initial_dist_sim_pairs_with_netcomp Public Method"
)
m = len(gs1)
n = len(gs2)
print("m: ", m, ", n: ", n)
dist_mat = np.zeros((m, n))
for i in range(m):
for j in range(n):
g1 = gs1[i]
g2 = gs2[j]
adj1, adj2 = [nx.adjacency_matrix(G) for G in [g1, g2]]
# print("nx.info(g1): ", nx.info(g1))
# print("nx.info(g2): ", nx.info(g2))
gid1 = g1.graph['gid']
gid2 = g2.graph['gid']
pair = (gid1, gid2)
# d = nc.lambda_dist(adj1,adj2,kind='laplacian',k=10)
d = nc.lambda_dist(adj1, adj2, kind='adjacency', k=10)
self.gidpair_ds_map[pair] = d
print('Adding entry ({}, {}) to dist map'.format(pair, d))
save(self.sfn, self.gidpair_ds_map)
print('Saved to {}'.format(self.sfn))
def lambda_dist(self, other_kg, **kwargs):
""" A NetComp operation returning the lambda distance between `self` and `other_kg`.
This operation is very complicated and is better described by NetComp's docs: https://github.com/peterewills/NetComp/blob/master/netcomp/distance/exact.py#L123
:param other_kg: The other knowledge graph
:type other_kg: :class:`.KnowledgeGraph`
:return: The lambda distance between `self` and `other_kg`
:rtype: float
"""
A, B = make_adjacency_matrices(self, other_kg)
return nc.lambda_dist(A, B, **kwargs)
def return_lambdas(k):
return [
lambda A1, A2: nc.lambda_dist(A1, A2, kind="adjacency", k=k),
lambda A1, A2: nc.lambda_dist(A1, A2, kind="laplacian", k=k),
lambda A1, A2: nc.lambda_dist(A1, A2, kind="laplacian_norm", k=k),
]
N = 100
M = 10\
G = nx.grid_2d_graph(N, M)
deg_seq = [item[1] for item in G.degree_iter()]
####################################
## DEFINE IMPORTANT FUNCTIONS
####################################
def distance(dist_func, A, B):
return dist_func(A, B)
lambda_adj = lambda A1, A2: nc.lambda_dist(A1, A2, kind="adjacency")
lambda_lap = lambda A1, A2: nc.lambda_dist(A1, A2, kind="laplacian")
lambda_nlap = lambda A1, A2: nc.lambda_dist(A1, A2, kind="laplacian_norm")
res_dist = lambda A1, A2: nc.resistance_distance(A1, A2, check_connected=False)
distances = [
nc.edit_distance,
res_dist,
nc.deltacon0,
nc.netsimile,
lambda_adj,
lambda_lap,
lambda_nlap,
]
labels = [
"Edit",
num_cores = multiprocessing.cpu_count()
# size of ensemble
ensemble_len = 500
threshold = 0.5
ones = True
####################################
## DEFINE IMPORTANT FUNCTIONS
####################################
def distance(dist_func,A,B): return dist_func(A,B)
lambda_adj = lambda A1,A2: nc.lambda_dist(A1,A2,kind='adjacency')
lambda_lap = lambda A1,A2: nc.lambda_dist(A1,A2,kind='laplacian')
lambda_nlap = lambda A1,A2: nc.lambda_dist(A1,A2,kind='laplacian_norm')
lambda_adj_5 = lambda A1,A2: nc.lambda_dist(A1,A2,kind='adjacency')
lambda_lap_5 = lambda A1,A2: nc.lambda_dist(A1,A2,kind='laplacian')
lambda_nlap_5 = lambda A1,A2: nc.lambda_dist(A1,A2,kind='laplacian_norm')
res_dist = lambda A1,A2: nc.resistance_distance(A1,A2,check_connected=False)
distances = [nc.edit_distance,res_dist,nc.deltacon0,nc.netsimile,
lambda_adj,lambda_lap,lambda_nlap,lambda_adj_5,lambda_lap_5,
lambda_nlap_5]
labels = ['Edit','Resistance Dist.','DeltaCon','NetSimile',
'Lambda (Adjacency)','Lambda (Laplacian)',
# A trivial example, for now import netcomp as nc import networkx as nx # use networkx to build a sample graph G0 = nx.complete_graph(10) G1 = G0.copy() G1.remove_edge(1, 0) G2 = G0.copy() G2.remove_node(0) A, B, C = [nx.adjacency_matrix(G) for G in [G0, G1, G2]] # matrix distances nc.resistance_distance(A, B) nc.edit_distance(A, B) nc.deltacon0(A, B) nc.vertex_edge_distance(A, B) # spectral distances nc.lambda_dist(A, C) nc.lambda_dist(A, C, kind='adjacency', k=2) nc.lambda_dist(A, C, kind='laplacian_norm') # other distances nc.resistance_distance(A, C, renormalized=True) nc.netsimile(A, C) # matrices w/o associated distances nc.commute_matrix(A) nc.conductance_matrix(A)