def test_features():
g = create_graph(2, GraphType=nx.DiGraph)
# print(attraction_basin(g))
# print(flow(g))
motif_level = 4
# with silence_stdout():
# example_feature(g)
# clustering_coefficient(g)
# k_core(g)
# node_page_rank(g)
# bfs_moments(g)
# print(m_res)
# gpu_motif = motif(g,level=motif_level,gpu=True)
# print(len(gpu_motif),'X',len(gpu_motif[0]))
# print(gpu_motif)
# print(gpu_motif == m_res)
# G = nx.random_regular_graph(20,100,seed=123456).to_directed()
G = nx.erdos_renyi_graph(30, 0.3, directed=True, seed=None)
# nx.write_gpickle(G,"test.pickle")
# G = g
# assert type(G) is nx.DiGraph
logger = PrintLogger("My Logger")
feature = MotifsNodeCalculator(G, level=motif_level, logger=logger)
feature.build()
mx = feature.to_matrix(mtype=np.matrix, should_zscore=False)
# print(mx[0])
m_res = motif(G, level=motif_level, gpu=False)
# pretify_motif_results(m_res)
gpu_res = motif(G, level=motif_level, gpu=True)
pretify_motif_results([m_res, gpu_res])
# pretify_motif_results([mx.tolist(),m_res,gpu_res])
m_res_arr = np.asarray(m_res)
gpu_res_arr = np.asarray(gpu_res)
print(mx.shape, m_res_arr.shape)
mx = gpu_res_arr
# np_diff = mx - m_res_arr
np_diff = mx - m_res_arr
print('diff avg:', np.average(np_diff[np.abs(np_diff) > 1]))
print('number of elements:', np.size(np_diff))
print('number of different elements:',
np.size(np_diff[np.abs(np_diff) > 1]))
print('Which is {}% of the elements'.format(
100 * np.size(np_diff[np.abs(np_diff) > 1]) / np.size(np_diff)))
list_diff = []
count = 0
for l in np_diff:
count += 1
list_diff.append(l.tolist())
# print(list_diff[count-1])
# list_diff = [l[0] for l in list_diff]
# print(list_diff)
# print('count',count)
pretify_motif_results([list_diff], lambda x: x != 0 and abs(x) > 1)
avg = (mx + m_res_arr) / 2
diff = (mx - m_res_arr)
a = diff[avg != 0] / avg[avg != 0]
a = a[np.abs(a) > 0.1]
print(a)
import os
import sys
import networkx as nx
import pickle
sys.path.append(os.path.abspath('.'))
sys.path.append(os.path.abspath('..'))
sys.path.append(os.path.abspath('../..'))
from src.accelerated_graph_features.feature_wrappers import motif
FEATURES = {
'Motif3': lambda g: motif(g, level=3, gpu=False),
'Motif4': lambda g: motif(g, level=4, gpu=False)
}
def calculate_features_on_graph(file_name, feats):
save_file_name = file_name + '_feature_results_{}.pickle'
g = nx.read_gpickle(file_name)
for k, func in feats.items():
print(k)
res = func(g)
with open(save_file_name.format(k), 'w+b') as fp:
pickle.dump(res, fp)
del res
def test_specific():
g = create_graph(3, GraphType=nx.DiGraph)
#g = nx.gnp_random_graph(50, 0.5, directed=True, seed=123456)
# print(attraction_basin(g))
print(motif(g, level=3, gpu=True))
motif, attraction_basin, flow
from src.accelerated_graph_features.graph_timer import FeatureTimer
from src.accelerated_graph_features.original_features_check import original_bfs_moments
# from Git.graph_measures.cpp_comparison import compare_motifs
from features_algorithms.vertices.attractor_basin import AttractorBasinCalculator
from features_algorithms.vertices.flow import FlowCalculator
import networkx as nx
matching_python_functions = {
# 'clustering': [clustering_coefficient, nx.average_clustering],
# 'k_core': [k_core, nx.k_core],
# 'page_rank': [node_page_rank, nx.pagerank]
# 'BFSMoments': [bfs_moments, original_bfs_moments]
# 'Motif3': [lambda G,timer: motif(G, level=3,timer=timer), lambda G: motif(G, level=3)],
# 'Motif4': [lambda G,timer: motif(G, level=4,timer=timer), lambda G: motif(G, level=4)]
'Motif3_GPU': {'cpp':lambda G,timer: motif(G, level=3,timer=timer),'gpu':lambda G,timer: motif(G, level=3,timer=timer,gpu=True)},
'Motif4_GPU': {'cpp':lambda G,timer: motif(G, level=4,timer=timer),'gpu':lambda G,timer: motif(G, level=4,timer=timer,gpu=True)}
# 'attraction_basin': [attraction_basin, lambda G: AttractorBasinCalculator(G).build()],
# 'flow': [flow, lambda G: FlowCalculator(G).build()]
}
SEED = 123456
def benchmark_gpu_feature_regular_graph(feature_name, type='cpp', d=20):
assert type in ['cpp', 'python', 'gpu']
nodes = [50, 100, 500, 1000, 2000, 5000, 10000, 100000]
if type is 'python':
titles = ['Feature calculation time']
else: