def make_feats(g):
counts5 = np.array(orca.orbit_counts("node", 5, g))
for v, n in zip(counts5, g.nodes):
if g.nodes[n]["node_feature"][0] > 0:
anchor_v = v
break
v5 = np.sum(counts5, axis=0)
return v5, anchor_v
def count_exact(queries, targets, args):
print("WARNING: orca only works for node anchored")
# TODO: non node anchored
n_matches_baseline = np.zeros(73)
for target in targets:
counts = np.array(orca.orbit_counts("node", 5, target))
if args.count_method == "bin":
counts = np.sign(counts)
counts = np.sum(counts, axis=0)
n_matches_baseline += counts
# don't include size < 5
n_matches_baseline = list(n_matches_baseline)[15:]
counts5 = []
num5 = 10 #len([q for q in queries if len(q) == 5])
for x in list(sorted(n_matches_baseline, reverse=True))[:num5]:
print(x)
counts5.append(x)
print("Average for size 5:", np.mean(np.log10(counts5)))
atlas = [
g for g in nx.graph_atlas_g()[1:] if nx.is_connected(g) and len(g) == 6
]
queries = []
for g in atlas:
for v in g.nodes:
g = g.copy()
nx.set_node_attributes(g, 0, name="anchor")
g.nodes[v]["anchor"] = 1
is_dup = False
for g2 in queries:
if nx.is_isomorphic(
g,
g2,
node_match=(lambda a, b: a["anchor"] == b["anchor"])
if args.node_anchored else None):
is_dup = True
break
if not is_dup:
queries.append(g)
print(len(queries))
n_matches_baseline = count_graphlets(queries,
targets,
n_workers=args.n_workers,
method=args.count_method,
node_anchored=args.node_anchored,
min_count=10000)
counts6 = []
num6 = 20 #len([q for q in queries if len(q) == 6])
for x in list(sorted(n_matches_baseline, reverse=True))[:num6]:
print(x)
counts6.append(x)
print("Average for size 6:", np.mean(np.log10(counts6)))
return counts5 + counts6
import orca
import networkx as nx
counts = orca.orbit_counts("node", 5, nx.gnm_random_graph(30, 100))
print(f'Number of nodes: {len(counts)}')
print(f'Number of graphlets: {len(counts[0])}')
print('Counts: ', counts)