def maketsk(args):
rrg_iter = args.pop("rrg_iter")
graphs = rg.make_graphs_static(**args)
if rrg_iter > 0:
graphs = rrg.rule_rand_graphs(graphs,
numgr=500 + EXPERIMENT_REPEATS,
iter=rrg_iter)[0]
return graphs
def test_rulerand():
import util.random_graphs as rg
import structout as so
import graphlearn3.util.setoperations as setop
grs = rg.make_graphs_static()[:30] #[:10] # default returns 100 graphs..
res1, grammar1 = rule_rand_graphs(grs, numgr=500, iter=2)
#res, grammar2=rule_rand_graphs(res1, numgr = 50, iter=1)
#so.gprint(res) #!!!!!!!!
'''
def test_neighexpansion():
graphs = rg.make_graphs_static(
7, # how many to generate
4, # graph size
3, # node-labelcount
2, # edgelabelcount
labeldistribution='uniform')
optimizer = pareto.LocalLandmarksDistanceOptimizer()
optimizer.enhance_grammar(graphs)
neighs = list(optimizer.grammar.neighbors(graphs[0]))
so.gprint(graphs[0])
so.gprint(neighs)
def test_randgraphs():
#make_graphs_static(n,ncnt,nlab,elab,maxdeg=3, labeldistribution='real'):
graphs = rg.make_graphs_static(
10, # how many to generate
5, # graph size
5, # node-labelcount
2, # edgelabelcount
labeldistribution='uniform')
so.graph.ginfo(graphs[0])
while graphs:
so.gprint(graphs[:3], edgelabel='label')
graphs = graphs[3:]
return graphs
def test_instancemaker():
graphs = rg.make_graphs_static(
7, # how many to generate
5, # graph size
5, # node-labelcount
2, # edgelabelcount
labeldistribution='uniform')
im = InstanceMaker(n_landmarks=3, n_neighbors=6).fit(graphs, ntargets=2)
landgraphs, des_dist, rest, target = im.get()
print("landmarks")
so.gprint(landgraphs, edgelabel='label')
print("des dist")
print(des_dist)
print("target")
so.gprint(target)
print("rest")
so.gprint(rest, edgelabel='label')
def test_grammar():
graphs = rg.make_graphs_static(
7, # how many to generate
5, # graph size
5, # node-labelcount
2, # edgelabelcount
labeldistribution='uniform')
optimizer = pareto.LocalLandmarksDistanceOptimizer()
optimizer.enhance_grammar(graphs)
print(optimizer.grammar)
keys = list(optimizer.grammar.productions.keys())
random.shuffle(keys)
print("start w grammar")
for k in keys[:10]:
cips = list(optimizer.grammar.productions[k].values())
so.gprint([c.graph for c in cips],
color=[[c.core_nodes, c.interface_nodes] for c in cips])
#so.graph.ginfo(cips[0].graph)
print(cips[0].__dict__)
def test_pareto():
configure_logging(logging.getLogger(), verbosity=2)
graphs = rg.make_graphs_static(
100, # how many to generate
5, # graph size
4, # node-labelcount
2, # edgelabelcount
labeldistribution='uniform',
allow_cycles=False)
im = InstanceMaker(n_landmarks=5, n_neighbors=50).fit(graphs, ntargets=2)
optimizer = pareto.LocalLandmarksDistanceOptimizer(n_iter=7,
context_size=1,
multiproc=True)
landmark_graphs, desired_distances, ranked_graphs, target_graph = im.get()
NONE = optimizer.optimize(
landmark_graphs,
desired_distances,
ranked_graphs,
#start_graph_list=[landmark_graphs[0]])
start_graph_list=landmark_graphs)
return None
