def km_random(g,k=5,m=3,start=None):
""" k nodes of breath first sequence; m add and del number."""
if start==None:
start=g.nodes().pop()
bfList=list(nx.bfs_edges(g,start))
bfList.reverse()
bfList.append((start,start))
tempk=[]
try:
while bfList:
for each in range(k):
tempk.append(bfList.pop()[1])
tg=nx.subgraph(g,tempk)
e=del_edge(tg,m)
g.remove_edges_from(e)
tg=nx.subgraph(g,tempk)
e=add_edge(tg,m)
g.add_edges_from(e)
tempk=[]
except IndexError:
print "pop finishing"
def calc_euler_tour(g, start, end):
'''Calculates an Euler tour over the graph g from vertex start to vertex end.
Assumes start and end are odd-degree vertices and that there are no other odd-degree
vertices.'''
even_g = nx.subgraph(g, g.nodes())
if end in even_g.neighbors(start):
# If start and end are neighbors, remove the edge
even_g.remove_edge(start, end)
comps = list(nx.connected_components(even_g))
# If the graph did not split, just find the euler circuit
if len(comps) == 1:
trail = list(nx.eulerian_circuit(even_g, start))
trail.append((start, end))
elif len(comps) == 2:
subg1 = nx.subgraph(even_g, comps[0])
subg2 = nx.subgraph(even_g, comps[1])
start_subg, end_subg = (subg1, subg2) if start in subg1.nodes() else (subg2, subg1)
trail = list(nx.eulerian_circuit(start_subg, start)) + [(start, end)] + list(nx.eulerian_circuit(end_subg, end))
else:
raise Exception('Unknown edge case with connected components of size {0}:\n{1}'.format(len(comps), comps))
else:
# If they are not neighbors, we add an imaginary edge and calculate the euler circuit
even_g.add_edge(start, end)
circ = list(nx.eulerian_circuit(even_g, start))
try:
trail_start = circ.index((start, end))
except:
trail_start = circ.index((end, start))
trail = circ[trail_start+1:] + circ[:trail_start]
return trail
def match_story_by_sen_edge(Gs, stories, target, tau):
existing = copy.deepcopy(stories['keywords_set'])
match_target = copy.deepcopy(target['keywords_set'])
node_cos = match_story(existing, match_target, 0.3)
subgs1 = []
for sto in match_target:
subgs1.append(nx.subgraph(Gs, sto))
subgs0 = []
for sto in existing:
subgs0.append(nx.subgraph(Gs, sto))
matched = []
for i in range(len(subgs1)):
matchingGraph = subgs1[i]
dis = []
for cand in subgs0:
val = compute_distance(matchingGraph, cand)
dis.append(val)
total = np.multiply(dis, node_cos[i])
match_score = np.max(total)
if match_score < tau:
stories['keywords_set'].append(target['keywords_set'][i])
stories['doc_set'].append(target['doc_set'][i])
continue
print match_score
match_ind = np.argmax(total)
print match_ind, stories['doc_set'][match_ind], target['doc_set'][i]
# match_text = existing[match_ind]
stories['keywords_set'][match_ind].extend(match_target[i])
u = stories['doc_set'][match_ind].union(target['doc_set'][i])
stories['doc_set'][match_ind] = u
def test_subgraph_of_subgraph(self):
SGv = nx.subgraph(self.G, range(3, 7))
SDGv = nx.subgraph(self.DG, range(3, 7))
SMGv = nx.subgraph(self.MG, range(3, 7))
SMDGv = nx.subgraph(self.MDG, range(3, 7))
for G in self.graphs + [SGv, SDGv, SMGv, SMDGv]:
SG = nx.induced_subgraph(G, [4, 5, 6])
assert_equal(list(SG), [4, 5, 6])
SSG = SG.subgraph([6, 7])
assert_equal(list(SSG), [6])
# subgraph-subgraph chain is short-cut in base class method
assert_is(SSG._graph, G)
def fitness(new_members, is_print=False):
if len(new_members) == 1:
return 0
else:
new_nodes = set(flatten(map(lambda mem: nx.neighbors(data_graph, mem), new_members))) | new_members
global w_in
global w_all
w_all = len(nx.subgraph(data_graph, new_nodes).edges())
w_in = len(nx.subgraph(data_graph, new_members).edges())
if is_print:
print 'w_in', w_in, nx.subgraph(data_graph, new_members).edges()
print 'w_all', w_all, nx.subgraph(data_graph, new_nodes).edges()
return float(w_in) / w_all
def test_subgraph(self):
assert_equal(self.G.subgraph([0, 1, 2, 4]).adj,
nx.subgraph(self.G, [0, 1, 2, 4]).adj)
assert_equal(self.DG.subgraph([0, 1, 2, 4]).adj,
nx.subgraph(self.DG, [0, 1, 2, 4]).adj)
assert_equal(self.G.subgraph([0, 1, 2, 4]).adj,
nx.induced_subgraph(self.G, [0, 1, 2, 4]).adj)
assert_equal(self.DG.subgraph([0, 1, 2, 4]).adj,
nx.induced_subgraph(self.DG, [0, 1, 2, 4]).adj)
# subgraph-subgraph chain is allowed in function interface
H = nx.induced_subgraph(self.G.subgraph([0, 1, 2, 4]), [0, 1, 4])
assert_is_not(H._graph, self.G)
assert_equal(H.adj, self.G.subgraph([0, 1, 4]).adj)
def filterGraphByRecipeID(G, Grecipes, Gingredients, recipeNodes): recipe_to_remove = [ n for n in Grecipes.nodes() if n not in recipeNodes] searchGrecipes = nx.subgraph(Grecipes, recipeNodes) searchGrecipes.remove_nodes_from(recipe_to_remove) ingrNodes = list(set([b for n in searchGrecipes.nodes() for b in G.neighbors(n)])) ingr_to_remove = [ n for n in Gingredients.nodes() if n not in ingrNodes] searchGingredients = Gingredients searchG = nx.subgraph(G, recipeNodes + ingrNodes) searchG.remove_nodes_from(recipe_to_remove) searchG.remove_nodes_from(ingr_to_remove) return (searchG, searchGrecipes, searchGingredients)
def plot_induced_subgraphs(self):
plt.figure(1)
partition = self.find_partition()[1]
communities = [partition[v] for v in partition]
newGraph=self.G
for community in communities:
nx.subgraph(newGraph, [key for key in partition if partition[key]==community])
node_color=[float(partition[v]) for v in partition]
labels = {}
for node in newGraph.nodes():
labels[node]= newGraph.node[node].get('name', '')
nx.draw_spring(newGraph,node_color=node_color, labels=labels)
plt.show()
plt.savefig("C:\\Users\\Heschoon\\Dropbox\\ULB\\Current trends of artificial intelligence\\Trends_project\\graphs\\graph_induced.pdf")
def core_substitution(graph, orig_cip_graph, new_cip_graph):
"""
graph is the whole graph..
subgraph is the interfaceregrion in that we will transplant
new_cip_graph which is the interface and the new core
"""
assert( set(orig_cip_graph.nodes()) - set(graph.nodes()) == set([]) ), 'orig_cip_graph not in graph'
# select only the interfaces of the cips
new_graph_interface_nodes = [n for n, d in new_cip_graph.nodes(data=True) if 'core' not in d]
new_cip_interface_graph = nx.subgraph(new_cip_graph, new_graph_interface_nodes)
original_graph_interface_nodes = [n for n, d in orig_cip_graph.nodes(data=True) if 'core' not in d]
original_interface_graph = nx.subgraph(orig_cip_graph, original_graph_interface_nodes)
# get isomorphism between interfaces, if none is found we return an empty graph
iso = get_good_isomorphism(graph,
orig_cip_graph,
new_cip_graph,
original_interface_graph,
new_cip_interface_graph)
if len(iso) != len(original_interface_graph):
# print iso
# draw.display(orig_cip_graph)
# draw.display(new_cip_graph)
#draw.graphlearn([orig_cip_graph, new_cip_graph],size=10)
logger.log(5,"grammar hash collision, discovered in 'core_substution' ")
return nx.Graph()
# ok we got an isomorphism so lets do the merging
graph = nx.union(graph, new_cip_graph, rename=('', '-'))
# removing old core
# original_graph_core_nodes = [n for n, d in orig_cip_graph.nodes(data=True) if 'core' in d]
original_graph_core_nodes = [n for n, d in orig_cip_graph.nodes(data=True) if 'core' in d]
for n in original_graph_core_nodes:
graph.remove_node(str(n))
# merge interfaces
for k, v in iso.iteritems():
graph.node[str(k)][
'interface'] = True # i am marking the interface only for the backflow probability calculation in graphlearn, this is probably deleteable because we also do this in merge, also this line is superlong Ooo
merge(graph, str(k), '-' + str(v))
# unionizing killed my labels so we need to relabel
return nx.convert_node_labels_to_integers(graph)
def node_coordinates_robust(graph):
xs, ys, zs, As, namess = [], [], [], [], []
for cc in nx.connected_components(graph.to_undirected()):
if len(cc) == 1:
x, y, z = [0], [0], [0]
A = np.array([[0]])
names = list(cc)
elif len(cc) == 2:
x, y, z = [0, 1], [0, 1], [0, 1]
n1, n2 = list(cc)
A = np.array([[0, 1], [1, 0]]) * graph[n1][n2].get('weight', 1)
names = list(cc)
else:
x, y, z, A, names = node_coordinates(nx.subgraph(graph, cc))
xs.append(x)
ys.append(y)
zs.append(z)
As.append(A)
namess.append(names)
for coord in [xs, ys, zs]:
loc = 0
for i, arr in enumerate(coord):
arr = np.asanyarray(arr)
scale = np.sqrt(arr.size - 0.99)
coord[i] = ((arr - np.min(arr)) / np.max(arr) *
scale + loc)
loc += 1.05 * scale
x = np.concatenate(xs)
y = np.concatenate(ys)
z = np.concatenate(zs)
A = coo_mat_concat(As)
names = list(itertools.chain(*namess))
return x, y, z, A, names
def congressSubgraph(base_graph, title='Rep'):
"""Generate a subgraph from a Congerssional Twitter graph based on a member's title"""
header = ['lastname','firstname','middlename','title','party','state','gender','twitter_id']
congress_reader = csv.DictReader(open('../twitter_congress.csv', 'rU'), fieldnames=header)
all_congress = map(lambda l: l, congress_reader)
mem_list = [(a['twitter_id']) for (a) in all_congress if a['title']==title]
return nx.subgraph(base_graph, mem_list)
def test_msf_components():
grid, dist_matrix = random_settlements(500)
msf = mod_boruvka(grid)
msf_subgraph = lambda components: nx.subgraph(msf, components)
component_graphs = map(msf_subgraph, nx.connected_components(msf))
def full_graph(g):
new_graph = nx.Graph()
new_graph.add_nodes_from(g.nodes(data=True))
if len(g.nodes()) < 2:
return new_graph
new_graph.add_weighted_edges_from([(u, v, dist_matrix[u][v])
for u, v in itertools.product(g.nodes(), g.nodes())
if u != v])
return new_graph
full_graphs = map(full_graph, component_graphs)
mst_graphs = map(nx.mst.minimum_spanning_tree, full_graphs)
diff_component_mst = []
for i in range(len(component_graphs)):
c_sets = set([frozenset(e) for e in component_graphs[i].edges()])
mst_sets = set([frozenset(e) for e in mst_graphs[i].edges()])
if not c_sets == mst_sets:
diff_component_mst.append(i)
assert len(diff_component_mst) == 0, str(len(diff_component_mst)) + \
" components are not MSTs"
def step3_balance_number(G,node): right=0 wrong=0 total=0 cate_num=dict() #the number 16 cases ins=find_in_nodes(G,node) outs=find_out_nodes(G,node) sub_nodes=ins+outs subG=nx.subgraph(G,sub_nodes) for in_node in ins: c_outs=find_in_nodes(subG,in_node) for out_node in c_outs: if out_node in outs: total=total+1 #node-->out_node-->mid_node-->in_node-->node flag1=(G[node][out_node]['weight']==-1) flag2=(G[out_node][in_node]['weight']==-1) flag3=(G[in_node][node]['weight']==-1) key=str(1000+flag1*100+flag2*10+flag3) if cate_num.has_key(key): cate_num[key]=cate_num[key]+1 else: cate_num[key]=1 if((flag1+flag2+flag3)%2==0): right=right+1 else: wrong=wrong+1 return (right,total,cate_num)
def init():
global projectname
global version_aray
global pos
global x
global y
global size_array
global numframes
global sg
for i in range(6):
data_directory = projectname + "_history/" + projectname + version_array[i] + "/" + projectname
[g, lines] = creategraph.readfile(data_directory)
if i == 0:
sg = creategraph.refine(g, 45)
[pos, x, y] = creategraph.coordinate(sg)
size = creategraph.point_sizes(sg, lines)
zeros = np.array([0] * len(size))
print 'len(size) = ', len(size)
print 'zeros = ', zeros
size_array.append(zeros)
size_array.append(size)
else:
# create the graph induced by nodes from sg
subg = nx.subgraph(g, nx.nodes(sg))
print subg, sg
if nx.number_of_nodes(subg) != nx.number_of_nodes(sg):
print 'panic at 34'
else: # v this seems to be a error, but not
size = creategraph.point_sizes(sg, lines)
size_array.append(size)
x = np.array(x)
y = np.array(y)
size_array = np.array(size_array)
def subgraph_from_pathways(pathways, g):
"""
returns a topological annotated graph containing given pathways
"""
nodes = []
if type(pathways) == list:
for p in pathways:
nodes += nodes_from_pathway(p, g)
elif type(pathways) == str:
nodes += nodes_from_pathway(pathways, g)
h = nx.subgraph(g, nodes)
remove_edges = []
if type(pathways) == list:
for p in pathways:
for e in h.edges():
if not p in h.get_edge_data(*e)['pathways']:
remove_edges.append(e)
elif type(pathways) == str:
for e in h.edges():
if not pathways in h.get_edge_data(*e)['pathways']:
remove_edges.append(e)
for r in remove_edges:
h.remove_edge(*r)
return topological_annotate(h)
def greedy_choice(G, candidate, blue_nodes, black_nodes, visited):
"""Helper function to greedy cut"""
G.node[candidate][gc.PARTITION] = gc.BLUE
blue_cut_val = gc.cut_edges(nx.subgraph(G, visited))
G.node[candidate][gc.PARTITION] = gc.BLACK
black_cut_val = gc.cut_edges(nx.subgraph(G, visited))
if blue_cut_val > black_cut_val:
G.node[candidate][gc.PARTITION] = gc.BLUE
blue_nodes.add(candidate)
else:
black_nodes.add(candidate)
return blue_nodes, black_nodes
def remove_bridges(self, in_file_, start_id, delim_):
reader = csv.reader(open(in_file_), delimiter=delim_)
for line in reader:
self.G.remove_edge(int(line[0]) - start_id,int(line[1]) - start_id)
print "no of components after removing bridges: %d" % nx.number_connected_components(self.G)
comps = nx.connected_components(self.G)
for comp in comps:
print len(comp)
bfs = self.BreadthFirstLevels(1,100)
nbunch = [1]
for n in bfs:
#print(n)
val_ = n.values()
for set_ in val_:
nbunch += list(set_)
#print nbunch
print "start creating the induced graph!"
induced_g = nx.subgraph(self.G, nbunch)
self.G.clear()
# start_ = 0
# for n_ in induced_g:
# self.maps_[n_] = start_
# start_ += 1
# for n_1 in induced_g:
# for n_2 in induced_g:
# if n_1 in induced_g.neighbors(n_2):
# self.G.add_edge(maps_[n_1],maps_[n_2])
self.n = nx.number_of_nodes(induced_g)
self.G = induced_g
print "no of node: %d and no of edges: %d in induce graph!" % (self.G.number_of_nodes(), self.G.number_of_edges())
def compute_global_utility(graph):
"""
Return an index that quantifies how big the size of adopter
clusters is in the entire population of consumers. We call this
index 'Global utility' in our article.
This index computes the cluster-size-weighted average of adopter
clusters divided by the total number of consumers
So it goes from 0 to 1 and it's always increasing.
"""
N = len(graph.nodes())
adopters = get_adopters(graph)
clusters = nx.subgraph(graph, adopters)
cluster_sizes = [len(c) for c in nx.connected_components(clusters) if len(c) > 1]
if cluster_sizes:
# The weight of each cluster depends on its size
weights = np.array(cluster_sizes) / N
# Compute the weighted average
weigthed_average = np.average(cluster_sizes, weights=weights)
# Since the index needs to go between 0 and 1, we need to divide between N
# again
utility = weigthed_average / N
return utility
else:
return 0
def expand(seed_set):
members = seed_set
print 'seed:', members, nx.subgraph(data_graph, set(
flatten(map(lambda mem: nx.neighbors(data_graph, mem), members))) | members).edges()
is_change = True
while is_change:
to_check_neighbors = list(flatten(map(lambda mem: nx.neighbors(data_graph, mem), members)))
random.shuffle(to_check_neighbors)
print to_check_neighbors
is_change = False
# for neighbor in to_check_neighbors:
for neighbor in to_check_neighbors:
if fitness(members | {neighbor}) > fitness(members):
is_change = True
members.add(neighbor)
fitness(members, is_print=True)
print 'add neighbor:', neighbor, members, 'w_in:', w_in, 'w_all:', w_all
break
for member in members:
if fitness(members - {member}) > fitness(members):
is_change = True
members.remove(member)
fitness(members, is_print=True)
print 'remove member:', member, members, 'w_in', w_in, 'w_all:', w_all
break
print set(members)
print '\n----------------------------\n'
def main():
"""
Pre-processing:
load data, compute centrality measures, write files with node data
"""
print(nx.__version__)
# Load network data, create storage dict, and extract main component
depends=nx.read_edgelist("data/depends.csv",delimiter=",",create_using=nx.DiGraph(),nodetype=str,data=(("weight",time_from_today),))
depends.name="depends"
suggests=nx.read_edgelist("data/suggests.csv",delimiter=",",create_using=nx.DiGraph(),nodetype=str,data=(("weight",time_from_today),))
suggests.name="suggests"
imports=nx.read_edgelist("data/imports.csv",delimiter=",",create_using=nx.DiGraph(),nodetype=str,data=(("weight",time_from_today),))
imports.name="imports"
nets_dict={"depends":depends,"suggests":suggests,"imports":imports}
for k in nets_dict.keys():
main_component=nx.connected_component_subgraphs(nets_dict[k].to_undirected())[0].nodes()
nets_dict[k]=nx.subgraph(nets_dict[k],main_component)
# Run multiple measures on graphs and normalize weights
measure_list=[nx.in_degree_centrality,nx.betweenness_centrality,nx.pagerank]
for g in nets_dict.values():
multiple_measures(g,measure_list)
normalize_weights(g)
# Output networks in GraphML format (to store node attributes)
for i in nets_dict.items():
# print(i[1].edges(data=True))
nx.write_graphml(i[1],"data/"+i[0]+"_data.graphml")
print("")
print("All files written with data")
"""Visualization:
def weak_link_distribution(bn,N_clus=2,mcs=0,n_jumps=1):
jumps=[j[0] for j in networkb.find_th_jumps(bn,N_clus)]
jumps=sorted(list(set(jumps)),reverse=True)
if len(jumps)==0:
return []
pcs=jumps[0:min(n_jumps,len(jumps)-1)]
d=[]
for pc in pcs:
G=bn.get_Graph(pc,correlation='positive')
cluster_list=[x for x in networkx.connected_components(G) if x>mcs]
if len(cluster_list)<2:
continue
thmin=(jumps[jumps.index(pc)+1]+pc)/2
print thmin
print pc
H=bn.get_Graph(thmin,th_up=pc,correlation='positive')
if H.number_of_edges()<1:
continue
H=networkx.subgraph(H,itertools.chain.from_iterable(cluster_list))
if H.number_of_edges()<1:
continue
for e in H.edges_iter():
d.append(bn.nodedistance(e))
json.dump(d,open(bn.weak_link_distribution_file,'w'))
return d
def filter_relevant_ids(graph):
"""
Dado el grafo de mis followed y sus followed,
extraemos los 100 nodos más relevantes
"""
graph = nx.read_gpickle('graph.gpickle')
my_followed = list(set([x[0] for x in graph.edges()]))
graph = nx.subgraph(graph, my_followed)
def get_nfollowed(nid):
return len(graph.successors(nid))
def get_nfollowers(nid):
return len(graph.predecessors(nid))
import pandas as pd
df = pd.DataFrame()
df['nodeid'] = my_followed
df['nfollowed'] = df['nodeid'].apply(get_nfollowed)
df['nfollowers'] = df['nodeid'].apply(get_nfollowers)
relevant = df[(df.nfollowed > 40) & (df.nfollowers > 40)]
relevantids = list(relevant.nodeid.values)
with open('layer0.pickle','wb') as f:
pickle.dump(relevantids, f)
return relevantids
def upper_bound(G, k, H):
'''
the upper bound of size of k-plex of H in G
'''
# upper_bound_by_deg = min([nx.degree(G, node) for node in H]) + k
# upper_bound_by_deg = 100
subg = nx.subgraph(G, H)
validate_neighbors = {node for node in neighbors(G, H) if len(set(G.neighbors(node)).intersection(H)) >= len(H)+1-k}
strict_nodes = [node for node in H if len(subg[node]) == len(H)-k]
if len(strict_nodes) > 0:
avaliable_nodes = {node for node in G.neighbors(strict_nodes[0]) if node not in H}
for i in range(1, len(strict_nodes)):
avaliable_nodes.intersection_update({node for node in G.neighbors(strict_nodes[i]) if node not in H})
avaliable_nodes.intersection_update(validate_neighbors)
return len(H)+len(avaliable_nodes)
else:
min_d = float('inf')
for node in H:
nbrs = neighbors(G, [node])
nbrs.intersection_update(validate_neighbors)
num_non_nbrs = k-1-(subg.number_of_nodes() - nx.degree(subg, node))
min_d = min(min_d, len(nbrs)+num_non_nbrs)
return min_d
def find_football_communities():
""" Finds the communities produced for the football network, uses compare
methods to graph
"""
fgraph = CD.football_graph()
known = CD.football_known_c()
temp7 = known[7]
temp8 = known[8]
temp9 = known[9]
known[7] = temp8
known[8] = temp9
known[9] = temp7
center_g = nx.Graph()
center_g.add_nodes_from(range(12))
centers = nx.circular_layout(center_g, scale = 10)
pos = {}
subgraphs = [nx.subgraph(fgraph, c) for c in known]
count = -1
for g in subgraphs:
count += 1
(off_x, off_y) = centers[count]
pos_local = nx.circular_layout(g, scale=2.)
for n, place in pos_local.iteritems():
pos[n] = place + np.array([off_x, off_y])
compare_methods(fgraph,
'football_',
param=[1., 1., 5./115., 4, 0, .7, 20],
known=known,
pos=pos,
color_map={76:1, 11:2, 7:3, 102:4, 104:5, 47:6, 98:7,
96:8, 23:9, 94:10, 27:0},
data_path="FootballGames/football_metis")
def bound_branch(G, k ,q_nodes, is_use_cores=False, select_method='rand'):
'''
wrapper of branch and bound method
'''
ts = time.time()
global optimal
optimal = set()
k_neighbors = k_hop_nbrs_n(G, k, q_nodes)
sub = set(q_nodes)
sub.update(k_neighbors)
g = nx.subgraph(G, sub)
if is_use_cores:
cores = nx.core_number(g)
else:
cores = None
# print('subgraph ', g.nodes())
print('minimum degree of subgraph', minimum_degree(g))
print('k neighbors', len(k_neighbors))
BB(g, k, q_nodes, set(), cores, select_method)
print('the solution is', optimal)
te = time.time()
texe = round(te-ts, 2) # the execution time
return texe
def mean_geodesic(pg, debug=0):
"""
mean_geodesic() calculates the mean geodesic (shortest) distance
between two vertices in a network.
"""
length_sum = 0
if networkx.is_directed_acyclic_graph(pg):
n_pairs_with_paths = 0
else:
n_pairs_with_paths = ( pg.order() * ( pg.order() + 1 ) ) / 2
tg = networkx.subgraph(pg, pg.nodes())
for u in pg.nodes_iter():
tg.delete_node(u)
for v in tg.nodes_iter():
try:
length = networkx.shortest_path_length(pg,u,v)
if length > 0:
length_sum = length_sum + length
if networkx.is_directed_acyclic_graph(pg):
n_pairs_with_paths = n_pairs_with_paths + 1
except networkx.exception.NetworkXError:
pass
try:
geodesic = float(length_sum) / float(n_pairs_with_paths)
except:
geodesic = -999.
if debug:
print 'length_sum:\t', length_sum
print 'n_pairs_with_paths:\t', n_pairs_with_paths
return geodesic
def vis_coauthor_communities(graph, source, i, prefix, options, radius, overlap):
""" Finds the communities produced by different methods for the astro
citation network
"""
interest = CD.get_ball(graph, options[source][i], radius)
print "Displaying and computing for a subset of ", len(interest), " nodes."
sgraph = nx.subgraph(graph, interest)
cleaned = {}
for key in options.keys():
""" for generating sub community structure
"""
if key == source:
# split the overarching with the substructure
cleaned[source] = [options[source][i]]
options['Parallel Subcommunities'] = options[source][:i]
options['Parallel Subcommunities'].extend(options[source][i+1:])
key = 'Parallel Subcommunities'
filtered = [filter(lambda n: n in interest, c) for c in options[key]]
filtered = filter(lambda c: len(c) > 0, filtered)
cleaned[key] = filtered
cleaned[key] = CD.clean_of_duplicate_c(cleaned[key], overlap=overlap)
compare_methods(sgraph, prefix, options=cleaned)
def dyad_census(pg, debug=0, debuglog=0):
"""
dyad_census() calculates the number of null, asymmetric, and
mutual edges between all pairs of nodes in a directed graph.
"""
if not networkx.is_directed_acyclic_graph(pg):
logging.error('pyp_network.dyad_census() requires a directed graph as input!')
return 0
else:
census = {}
census['null'] = 0
census['asymmetric'] = 0
census['mutual'] = 0
tg = networkx.subgraph(pg, pg.nodes())
for u in pg.nodes_iter():
tg.delete_node(u)
for v in tg.nodes_iter():
if not pg.has_neighbor(u,v):
census['null'] = census['null'] + 1
elif u in pg.predecessors(v) and v in pg.successors(u):
census['mutual'] = census['mutual'] + 1
if debug:
print 'Nodes %s and %s link to one another!' % ( u, v )
if debuglog:
logging.error('Nodes %s and %s link to one another!',u, v)
elif u in pg.predecessors(v) and v not in pg.successors(u):
census['asymmetric'] = census['asymmetric'] + 1
elif u not in pg.predecessors(v) and v in pg.successors(u):
census['asymmetric'] = census['asymmetric'] + 1
else:
pass
del(tg)
return census
def network_properties(network : nx.DiGraph,
in_degree_threshold : float = -1,
pagerank_threshold : float = -1,
damping : float = 0.85,
spectral_offset : float = 0.5)\
-> (pd.DataFrame, sparse.spmatrix):
conn = max(nx.connected_components(network.to_undirected()), key=len)
conn = nx.subgraph(network, conn)
pr = compute_pagerank(conn, damping=damping)
names = nx.nodes(conn)
indeg = [conn.in_degree(n) for n in names]
odeg = [conn.out_degree(n) for n in names]
description = [conn.node[n].get('description', n) for n in names]
x, y, z, Adj, aff_names = node_coordinates(conn, nodelist=names,
offset=spectral_offset)
data = {'id': names,
'in_degree': indeg,
'out_degree': odeg,
'pagerank': pr,
'affinity_x': x,
'affinity_y': y,
'processing_depth': z,
'description': description}
df = pd.DataFrame(data, index=names)
df = df[df['pagerank'] > pagerank_threshold / len(names)]
df = df[df['in_degree'] > in_degree_threshold]
return df, Adj
def get_all_hardware_grid_problems(
device_graph: nx.Graph,
central_qubit: cirq.GridQubit,
n_instances: int,
rs: np.random.RandomState,
):
"""Helper function to get all subgraphs for a given named device.
This is annotated with lru_cache so you can freely call this function
multiple times without re-constructing the list of qubits.
Used by `generate_hardware_problem_problem` to get a subgraph for a given
value of n_qubits.
Returns:
A dictionary indexed by n_qubit, instance_i
"""
all_hg_problems: Dict[Tuple[int, int], HardwareGridProblem] = {}
subgraphs = get_growing_subgraphs(device_graph=device_graph,
central_qubit=central_qubit)
for n_qubits in sorted(subgraphs):
subgraph = nx.subgraph(device_graph, subgraphs[n_qubits])
for instance_i in range(n_instances):
problem = random_plus_minus_1_weights(subgraph, rs=rs)
qubits = sorted(problem.nodes)
coordinates = [(q.row, q.col) for q in qubits]
problem = nx.relabel_nodes(problem,
{q: i
for i, q in enumerate(qubits)})
all_hg_problems[n_qubits, instance_i] = HardwareGridProblem(
graph=problem,
coordinates=coordinates,
)
return all_hg_problems
def _build_list(self):
print "using disconnectivity analysis to find minima to untrap"
self.minpairs = deque()
graph = TSGraph(self.database).graph
cclist = list(nx.connected_components(graph))
# get the largest cluster
group1 = cclist[0]
min1 = sorted(group1, key=lambda m: m.energy)[0]
if not min1 == self.database.minima()[0]:
# make sure that the global minimum is in group1
print "warning, the global minimum is not the in the largest cluster."
# compute the energy barriers for all minima in the cluster
subgraph = nx.subgraph(graph, group1)
energy_barriers = self._compute_barriers(subgraph, min1)
# sort the minima by the barrier height divided by the energy difference
weights = [(m, np.abs(barrier) / np.abs(m.energy - min1.energy))
for (m, barrier) in energy_barriers.iteritems()]
weights.sort(key=lambda v: 1. / v[1])
self.minpairs = deque()
for min2, w in weights:
if len(self.minpairs) > self.list_len:
break
if not self.is_good_pair(min1, min2):
continue
self.minpairs.append((min1, min2))
if True:
# print some stuff
print " untrap analysis: minimum", min2.id(), "with energy", min2.energy, "barrier", energy_barriers[
min2], "untrap weight", w
def __HeadTailCommunityDetection(G, finaledgelist, head_tail_ratio=0.6):
H = nx.connected_components(G)
for s in H:
subgraph = nx.subgraph(G, s)
result = nx.edge_betweenness(subgraph, normalized=False)
edges = list(result.keys())
values = list(result.values())
mean = np.mean(values)
edgelist = []
edgetemp = subgraph.edges()
if len(edgetemp) <= 2:
for edge in edgetemp:
finaledgelist.append(edge)
else:
for index in range(len(values)):
if values[index] <= mean:
edgelist.append(edges[index])
if float(len(edgelist)) / float(
len(edges)
) <= head_tail_ratio: # change the head/tail division rule here, here is for tail percentage,
# so if the rule is 40/60, the value should be assigned 0.6 as in the code.
for edge in edgelist:
finaledgelist.append(edge)
else:
Gsub = nx.Graph()
for edge in edgelist:
Gsub.add_edge(edge[0], edge[1])
try:
__HeadTailCommunityDetection(Gsub, finaledgelist,
head_tail_ratio)
except:
pass
return finaledgelist
def avg_distance(graph: nx.Graph, communities: object,
**kwargs: dict) -> object:
"""Average distance.
The average distance of a community is defined average path length across all possible pair of nodes composing it.
:param graph: a networkx/igraph object
:param communities: NodeClustering object
:param summary: boolean. If **True** it is returned an aggregated score for the partition is returned, otherwise individual-community ones. Default **True**.
:return: If **summary==True** a FitnessResult object, otherwise a list of floats.
Example:
>>> from cdlib.algorithms import louvain
>>> from cdlib import evaluation
>>> g = nx.karate_club_graph()
>>> communities = louvain(g)
>>> scd = evaluation.avg_distance(g,communities)
"""
return __quality_indexes(
graph, communities,
lambda graph, coms: nx.average_shortest_path_length(
nx.subgraph(graph, coms)), **kwargs)
def is_consistent(queue, G, reactions, trail):
"""Checks whether the trail also makes sense when between a metabolic
pathway and the original gene neighborhood G.
:param queue: multiprocessing queue for storing the result of the procedure
:param G: undirected graph representing a genome, with genes
for vertices (i.e. not reaction identifiers as are vertices in D).
:param reactions: dict of dicts storing reaction information (obtained by
parsing a KGML file)
:param trail: CoMetGeNe trail
:return: True if trail is consistent with respect to the gene neighborhood
graph G, False otherwise
"""
G_nodes = G.nodes()
gene_trail = list()
for vertex in trail:
gene_vertex = list()
for gene in reactions[vertex]['enzyme']:
if gene in G_nodes:
gene_vertex.append(gene)
gene_trail.append(gene_vertex)
decompositions = decompose(gene_trail)
for decomp in decompositions:
for X in nx.connected_components(nx.subgraph(G, decomp)):
skip = False
for vertex in trail:
if not skip:
enzyme_names = reactions[vertex]['enzyme']
if len(set(X) & set(enzyme_names)) == 0:
skip = True
if not skip:
queue.put(True)
return
queue.put(False)
def parse_graph_from_csv(file, largest_cc=False, skip_first_line=True):
"""
Parses a file from a SNAP Dataset and returns a networkx graph.
Parameters
----------
file : file
largest_cc : bool
Whether the complete graph in the dataset should be returned, or only
its largest connected component.
"""
graph_name = os.path.splitext(os.path.basename(file.name))[0]
if skip_first_line:
file.readline() # ignore first line `node_1,node_2`
graph = nx.read_adjlist(file, delimiter=",", nodetype=int)
# Change indices to start from zero if necessary
if min(graph.nodes()) == 1:
# print("Reducing indices by one to have them starting at zero.")
graph = nx.relabel_nodes(graph, lambda x: x - 1, copy=False)
if largest_cc:
original_size = len(graph)
largest_component = max(nx.connected_components(graph), key=len)
subgraph = nx.subgraph(graph, largest_component)
graph = nx.Graph(subgraph)
if len(graph) < original_size:
print(
"Only considering largest connected component with %d nodes. Original graph had %d nodes."
% (len(graph), original_size))
else:
print("Graph has only one connected component.")
return graph_name, graph
def has_skipped_vertices_G(trail, reactions, G):
"""Determines whether the given CoMetGene trail skips any genes.
:param trail: CoMetGene trail
:param reactions: dict of dicts storing reaction information (obtained by
parsing a KGML file)
:param G: undirected graph built on the same vertex set as the metabolic
pathway (see Model in the methods section of Zaharia et al., 2018)
:return: True if the CoMetGeNe trail was obtained skipping at least one
gene, False otherwise
"""
involved = set() # genes involved in this trail
for r_id in trail:
for gene in reactions[r_id]['enzyme']:
involved.add(gene)
G_sub = nx.subgraph(G, trail)
for v1, v2 in G_sub.edges():
if 'skipped' in G_sub[v1][v2]:
for vertex in G_sub[v1][v2]['skipped']:
if vertex not in involved:
return True
return False
def entire_workflow(m, steps, evaluate_function, weight_function):
grid = nx.grid_graph([m, m])
for v in grid.nodes():
grid.node[v]["X"] = v[0]
grid.node[v]["Y"] = v[1]
grid.node[v]["pos"] = [v[0], v[1]]
grid.graph["size"] = 0
grid.graph["steps"] = steps
path = random_walk(grid, steps, True)
boundary_nodes = []
for x in grid.nodes():
if grid.degree(x) <= 3:
boundary_nodes.append(x)
boundary = nx.subgraph(grid, boundary_nodes)
restricted_path = restrict_path(boundary_nodes, path)
weight_function(boundary)
series = create_time_series(boundary, restricted_path, evaluate_function)
return [series, restricted_path, boundary]
def erdos_renyi_modularity(graph, communities, **kwargs):
"""Erdos-Renyi modularity is a variation of the Newman-Girvan one.
It assumes that vertices in a network are connected randomly with a constant probability :math:`p`.
.. math:: Q(S) = \\frac{1}{m}\\sum_{c \\in S} (m_S − \\frac{mn_S(n_S −1)}{n(n−1)})
where :math:`m` is the number of graph edges, :math:`m_S` is the number of community edges, :math:`l_S` is the number of edges from nodes in S to nodes outside S.
:param graph: a networkx/igraph object
:param communities: NodeClustering object
:return: FitnessResult object
Example:
>>> from cdlib.algorithms import louvain
>>> from cdlib import evaluation
>>> g = nx.karate_club_graph()
>>> communities = louvain(g)
>>> mod = evaluation.erdos_renyi_modularity(g,communities)
:References:
1. Erdos, P., & Renyi, A. (1959). `On random graphs I. <https://gnunet.org/sites/default/files/Erd%C5%91s%20%26%20R%C3%A9nyi%20-%20On%20Random%20Graphs.pdf/>`_ Publ. Math. Debrecen, 6, 290-297.
"""
graph = convert_graph_formats(graph, nx.Graph)
m = graph.number_of_edges()
n = graph.number_of_nodes()
q = 0
for community in communities.communities:
c = nx.subgraph(graph, community)
mc = c.number_of_edges()
nc = c.number_of_nodes()
q += mc - (m * nc * (nc - 1)) / (n * (n - 1))
return FitnessResult(score=(1 / m) * q)
def __init__(self,
function,
n_nodes=None,
p_edges=None,
max_iter_time=60
):
if n_nodes is None:
self.n_nodes = [int(i) for i in [1e2, 3e2, 1e3]]
else:
self.n_nodes = n_nodes
if p_edges is None:
self.p_edges = [0.0005, 0.001]
else:
self.p_edges = p_edges
self.function = function
self.simulation_grid = self.expand_grid_local(self.n_nodes, self.p_edges)
self._simulation_graphs = [nx.subgraph(graph, nbunch=max(nx.connected_components(graph), key=len))
for graph in self.random_graph_set(self.n_nodes, self.p_edges)]
self.simulation_grid['n_edges'] = [graph.number_of_edges() for graph in self._simulation_graphs]
self.computing_time = list()
self.computing_time_df = self.simulation_grid.copy()
self.max_iter_time = max_iter_time
def avg_transitivity(graph: nx.Graph, communities: object,
**kwargs: dict) -> object:
"""Average transitivity.
The average transitivity of a community is defined the as the average clustering coefficient of its nodes w.r.t. their connection within the community itself.
:param graph: a networkx/igraph object
:param communities: NodeClustering object
:param summary: boolean. If **True** it is returned an aggregated score for the partition is returned, otherwise individual-community ones. Default **True**.
:return: If **summary==True** a FitnessResult object, otherwise a list of floats.
Example:
>>> from cdlib.algorithms import louvain
>>> from cdlib import evaluation
>>> g = nx.karate_club_graph()
>>> communities = louvain(g)
>>> scd = evaluation.avg_transitivity(g,communities)
"""
return __quality_indexes(
graph, communities,
lambda graph, coms: nx.average_clustering(nx.subgraph(graph, coms)),
**kwargs)
def hub_dominance(graph: nx.Graph, communities: object,
**kwargs: dict) -> object:
"""Hub dominance.
The hub dominance of a community is defined as the ratio of the degree of its most connected node w.r.t. the theoretically maximal degree within the community.
:param graph: a networkx/igraph object
:param communities: NodeClustering object
:param summary: boolean. If **True** it is returned an aggregated score for the partition is returned, otherwise individual-community ones. Default **True**.
:return: If **summary==True** a FitnessResult object, otherwise a list of floats.
Example:
>>> from cdlib.algorithms import louvain
>>> from cdlib import evaluation
>>> g = nx.karate_club_graph()
>>> communities = louvain(g)
>>> scd = evaluation.hub_dominance(g,communities)
"""
return __quality_indexes(
graph, communities, lambda graph, coms: max(
[x[1] for x in list(nx.degree(nx.subgraph(graph, coms)))]) /
(len(coms) - 1), **kwargs)
def canonical_order(graph, face):
'''
Outputs the coordinates of the nodes of the face in a canonical order
in particular, the first one is the lex-min.
You need to use the graph structure to make this work
'''
lex_sorted_nodes = sorted(face)
first_node = lex_sorted_nodes[0]
cycle_sorted_nodes = [first_node]
local_cycle = nx.subgraph(graph, face)
#Compute the second node locally based on angle orientation
v = first_node
locations = []
neighbor_list = list(local_cycle.neighbors(v))
for w in neighbor_list:
locations.append(graph.nodes[w]["pos"] - graph.nodes[v]["pos"])
angles = [float(np.arctan2(x[1], x[0])) for x in locations]
neighbor_list.sort(key=dict(zip(neighbor_list, angles)).get)
second_node = neighbor_list[0]
cycle_sorted_nodes.append(second_node)
##Now compute a canonical ordering of local_cycle, clockwise, starting
##from first_node
while len(cycle_sorted_nodes) < len(lex_sorted_nodes):
v = cycle_sorted_nodes[-1]
neighbor_list = list(local_cycle.neighbors(v))
neighbor_list.remove(cycle_sorted_nodes[-2])
cycle_sorted_nodes.append(neighbor_list[0])
return cycle_sorted_nodes
def _graphize_streets(self, highways, node_suffix=''):
self.logger.debug('Creating street skeleton graph...')
street_graph = nx.Graph()
for h in highways:
raw_ns = h.get_nodes()
for n1, n2 in _walk2(raw_ns):
# trimming streets out of box
if not ((self.box[0] < n1.lat < self.box[2]) and (self.box[1] < n1.lon < self.box[3])):
if not ((self.box[0] < n2.lat < self.box[2]) and (self.box[1] < n2.lon < self.box[3])):
continue
street_graph.add_node(str(n2.id) + node_suffix,
pos=asarray([n1.lon, n1.lat]).astype(float),
type='street')
street_graph.add_node(str(n1.id) + node_suffix,
pos=asarray([n2.lon, n2.lat]).astype(float),
type='street')
street_graph.add_edge(str(n2.id) + node_suffix, str(n1.id) + node_suffix, phases=3, type='street')
# it may be possible that the streets are not connected. In that case, we only take the biggest component.
# one could also choose to add nonexistent street edges or to find the minimum street path to connect
# everything with openstreetmaps (dangerous).
if not nx.is_connected(street_graph):
self.logger.warning('The original graph is not connected. Only the biggest component will be kept.')
subgraphs = [nx.subgraph(street_graph, n) for n in nx.connected_components(street_graph)]
sizes = {len(x.nodes): x for x in subgraphs}
biggest_sg = sizes[max(sizes.keys())]
killnodes = [n for n in street_graph.nodes if n not in biggest_sg.nodes]
street_graph.remove_nodes_from(killnodes)
return street_graph
def show_subgraphs(ldag, context="edge", load_cache=True, graph=None):
if graph is None:
graph = ldag.graph
l_buf = []
separator = "\n\n"
iterobj = sorted(nx.connected_components(graph.to_undirected()),
key=len,
reverse=True)
for sgid, nodes in enumerate(iterobj):
if len(nodes) == 1:
continue
l_graph_buf = ["Subgraph {0} ({1} nodes)".format(sgid, len(nodes))]
subg = nx.subgraph(graph, nodes)
for edge in remove_edge_duplication(subg.edges(), ldag, graph=graph):
msg = edge_view(edge,
ldag,
context=context,
load_cache=load_cache,
graph=graph)
l_graph_buf.append(msg)
l_buf.append("\n".join(l_graph_buf))
return separator.join(l_buf)
def caculateSST(G, fracs):
KSD_DD = [0] * len(fracs)
KSD_CCD = [0] * len(fracs)
ND_DD = [0] * len(fracs)
ND_CCD = [0] * len(fracs)
# SDD_DD = [0] * len(fracs)
# SDD_CCD = [0] * len(fracs)
reNodes = SST(G)
for j in range(len(fracs)):
tempNodes = []
for node in reNodes:
if len(tempNodes) >= fracs[j] * G.number_of_nodes():
break
tempNodes.append(node)
sampleG = nx.subgraph(G, tempNodes)
dd2 = DD(sampleG)
ccd2 = CCD(sampleG)
KSD_DD[j] = KSD(dd1, dd2)
KSD_CCD[j] = KSD(ccd1, ccd2)
ND_DD[j] = ND(dd1, dd2)
ND_CCD[j] = ND(ccd1, ccd2)
# SDD_DD[j] = SDD(dd1, dd2)
# SDD_CCD[j] = SDD(ccd1, ccd2)
return KSD_DD, KSD_CCD, ND_DD, ND_CCD
def init():
global max_depth
global T
global grid_dim_x
global grid_dim_y
global base
global route_length
global total_distance
global total_reward
total_distance = 0
total_reward = 0
max_depth = 2
T = 3
grid_dim_x = 26
grid_dim_y = 26
base = 0.0
route_length = 9000
global graph
global dist
global sigma1
global avg_strat1
global regret1
global route
df = pd.read_csv("data/paws_mdp_out.txt", sep=" ")
dist = pd.read_csv("data/dist.gop", sep=" ", header=None)
graph = \
nx.from_pandas_dataframe(df, source='node_from', target='node_to',
edge_attr=['distance', 'animal_density', 'grid_cell_x', 'grid_cell_y'])
graph = nx.subgraph(graph, nx.node_connected_component(graph, base))
sigma1 = [[1/(grid_dim_x*grid_dim_y)] * grid_dim_y] * grid_dim_x
avg_strat1 = [[0] * grid_dim_y] * grid_dim_x
regret1 = [[0] * grid_dim_y] * grid_dim_x
route = []
def get_stat_datasets():
# path = '/network/rit/lab/ceashpc/share_data/GraphOpt/datasets/DBLP/DBLP_Citation_2014_May/DM'
path = '/network/rit/lab/ceashpc/share_data/GraphOpt/datasets/DBLP/DBLP_Citation_2014_May/DB'
# fn = 'dm_top30000_dataset.pkl'
fn = 'db_top30000_dataset.pkl'
with open(os.path.join(path, fn), 'rb') as rfile:
dataset = pickle.load(rfile)
first_graph = dataset['first_graph']
second_graph = dataset['second_graph']
print(nx.density(first_graph))
print(nx.density(second_graph))
print(first_graph.number_of_nodes())
print(first_graph.number_of_edges())
print(second_graph.number_of_nodes())
print(second_graph.number_of_edges())
print(nx.is_connected(first_graph))
print(nx.is_connected(second_graph))
lcc = max(nx.connected_component_subgraphs(second_graph), key=len)
subgraph = nx.subgraph(second_graph, lcc)
print(subgraph.number_of_nodes())
print(subgraph.number_of_edges())
def set_splice_graph_old(self, sg, component, target):
self.graph = sg.get_graph()
self.tx_paths = sg.annotation
self.original_tx_paths = sg.annotation # tx paths all ways without trimming
known_edges = set([(tx[i], tx[i + 1]) for tx in self.tx_paths
for i in range(len(tx) - 1)])
self.component = component
self.target = target
self.sub_graph = nx.subgraph(self.graph, self.component)
# add any possible tx that uses novel edges to list of known txs
for tx in nx.all_simple_paths(self.sub_graph,
source=self.component[0],
target=self.component[-1]):
novel = False
for i in range(len(tx) - 1):
if (tx[i], tx[i + 1]) not in known_edges:
novel = True
if novel:
self.tx_paths.append(tx)
self.inc_lengths, self.skip_lengths = [], [
] # call set all_path_lengths method
self.all_path_coordinates = [] # call set_all_path_coordinates method
for its in range(1, 100):
et = [0, 0]
while et[0] == et[1]:
ed = random.choice(el)
et = [
int(df.loc[df["ID"] == ed[0], "newtree" + str(its - 1)]),
int(df.loc[df["ID"] == ed[1], "newtree" + str(its - 1)])
]
sgn = []
for n in graph.nodes():
if int(df.loc[df["ID"] == n, "newtree" + str(its - 1)]) in et:
sgn.append(n)
sgraph = nx.subgraph(graph, sgn)
edd = {0: et[0], 3: et[1]}
newtree_partial = recursive_tree_part(sgraph, 2, "POP10", .05, 2)
newtree = {}
for n in graph.nodes():
if n not in sgn:
newtree[n] = int(df.loc[df["ID"] == n, "newtree" + str(its - 1)])
else:
newtree[n] = edd[newtree_partial[n]]
df["newtree" + str(its)] = df["ID"].map(newtree)
#df["newtree"+str(its)]=pd.to_numeric(df["newtree"+str(its)])
df.plot(column="newtree" + str(its), cmap="tab20")
def setup(self, **keywords):
"""
2012.10.15
run before anything is run
"""
AbstractMatrixFileWalker.setup(self, **keywords)
#self.writer = BeagleGenotypeFile(path=self.outputFname, mode='w')
#read in the IBD check result
self.ibdData = SNP.readAdjacencyListDataIntoMatrix(inputFname=self.pedigreeKinshipFilePath, \
rowIDHeader=None, colIDHeader=None, \
rowIDIndex=0, colIDIndex=1, \
dataHeader=None, dataIndex=2, hasHeader=False)
#. read in the alignment coverage data
alignmentCoverageFile = MatrixFile(
path=self.individualAlignmentCoverageFname)
alignmentCoverageFile.constructColName2IndexFromHeader()
alignmentReadGroup2coverageLs = alignmentCoverageFile.constructDictionary(
keyColumnIndexList=[0], valueColumnIndexList=[1])
alignmentCoverageFile.close()
sys.stderr.write(
"Reading in all samples from %s VCF input files ... \n" %
(len(self.inputFnameLs)))
# read all the Beagle files
individualID2HaplotypeData = {}
for inputFname in self.inputFnameLs:
vcfFile = VCFFile(inputFname=inputFname)
#vcfFile.readInAllHaplotypes()
for individualID in vcfFile.getSampleIDList():
individualID2HaplotypeData[individualID] = None
#haplotypeList = vcfFile.getHaplotypeListOfOneSample(individualID)
#individualID2HaplotypeData[individualID] = PassingData(haplotypeList=haplotypeList,
# locusIDList=vcfFile.locusIDList)
# get all haplotypes , etc.
# get all sample IDs
sys.stderr.write("%s individuals total.\n" %
(len(individualID2HaplotypeData)))
#. read in the pedigree or deduce it from Beagle Trio/Duo genotype file (columns)
#. construct individualID2pedigreeContext, context: familySize=1/2/3, familyPosition=1/2 (parent/child)
sys.stderr.write("Constructing individualID2pedigreeContext ...")
plinkPedigreeFile = PlinkPedigreeFile(path=self.pedigreeFname)
pGraph = plinkPedigreeFile.pedigreeGraph
#shrink the graph to only individuals with data
pGraph = nx.subgraph(pGraph, individualID2HaplotypeData.keys())
cc_subgraph_list = nx.connected_component_subgraphs(
pGraph.to_undirected())
individualID2familyContext = {}
outDegreeContainer = NumberContainer(minValue=0)
familySizeContainer = NumberContainer(minValue=0)
individualCoverageContainer = NumberContainer(minValue=0)
familyCoverageContainer = NumberContainer(minValue=0)
for cc_subgraph in cc_subgraph_list:
familySize = len(cc_subgraph)
familySizeContainer.addOneValue(familySize)
familyCoverage = 0
for n in cc_subgraph: #assuming each family is a two-generation trio/nuclear family
individualCoverage = self.getIndividualCoverage(
individualID=n,
alignmentReadGroup2coverageLs=alignmentReadGroup2coverageLs
)
individualCoverage = float(individualCoverage)
individualCoverageContainer.addOneValue(individualCoverage)
familyCoverage += individualCoverage
in_degree = pGraph.in_degree(n)
out_degree = pGraph.out_degree(n)
outDegreeContainer.addOneValue(out_degree)
familyContext = PassingData(familySize=familySize, in_degree=in_degree, out_degree=out_degree, \
individualCoverage=individualCoverage,\
familyCoverage=None)
if n not in individualID2familyContext:
individualID2familyContext[n] = familyContext
else:
sys.stderr.write(
"Node %s already in individualID2familyContext.\n" %
(n))
familyCoverageContainer.addOneValue(familyCoverage)
#set the family coverage for each member, used in weighing the individual. better covered family => better haplotype
for n in cc_subgraph:
individualID2familyContext[n].familyCoverage = familyCoverage
plinkPedigreeFile.close()
sys.stderr.write("%s individuals.\n" %
(len(individualID2familyContext)))
# weigh each unique individual based on its sequencing coverage + no of offspring => probability mass for each individual
sys.stderr.write(
"Weighing each individual , assigning probability mass ...")
individualID2probabilityMass = {}
for individualID, familyContext in individualID2familyContext.items():
outDegreeQuotient = outDegreeContainer.normalizeValue(
familyContext.familySize)
individualCoverageQuotient = individualCoverageContainer.normalizeValue(
familyContext.individualCoverage)
#familyCoverageQuotient = familyCoverageContainer.normalizeValue(familyContext.familyCoverage)
importanceScore = outDegreeQuotient + individualCoverageQuotient
representativeImportanceScore = importanceScore
individualID2probabilityMass[
individualID] = representativeImportanceScore
sys.stderr.write(" %s IDs with probability mass assigned.\n" %
(len(individualID2probabilityMass)))
self.individualID2probabilityMass = individualID2probabilityMass
self.individualID2HaplotypeData = individualID2HaplotypeData
cache[fingerprint] = (mcost, mpath)
return (mcost, mpath)
main(None)
#wrapper(main)
import pygraphviz
from networkx.drawing.nx_agraph import write_dot
#print(G.edges(data=True))
#print(G.nodes)
#labels = nx.get_edge_attributes(G,'weight')
#nx.draw_networkx_edge_labels(G,pos=nx.spring_layout(G),edge_labels=labels)
cnt = 0
cache = dict()
write_dot(G, "maze.dot")
nx.draw_spring(G, with_labels=True)
plt.savefig("maze_nwx.png")
# cut the graph into the four distinct different graphs
graphs = list()
for i in range(0, 4):
graphs.append(nx.subgraph(G, nx.node_connected_component(G, "@" + str(i))))
cache = dict()
print(findtheway(graphs, ["@0", "@1", "@2", "@3"]))
def proj_mp(graph, weight, A, sparsity, lmbd, max_iter=10, epsilon=1e-3):
current_x = proj_init_point(graph.number_of_nodes()) + 1e-6
edges = np.array(graph.edges)
edge_weights = np.ones(graph.number_of_edges())
start_time = time.time()
for i in range(max_iter):
print('iter {}'.format(i))
iter_time = time.time()
gradient = proj_get_gradient(current_x, weight, A, lmbd)
normalized_gradient = normalize_gradient(current_x, gradient)
re_head = head_proj(edges=edges,
weights=edge_weights,
x=normalized_gradient,
g=1,
s=sparsity,
budget=sparsity - 1.,
delta=1. / 169.,
err_tol=1e-8,
max_iter=100,
root=-1,
pruning='strong',
epsilon=1e-10,
verbose=0)
re_nodes, _, _ = re_head
gamma_x = set(re_nodes)
print('gamma_x', len(gamma_x))
# print(sorted(list(gamma_x)))
if i == 0:
supp_x = set()
else:
supp_x = set(
[ind for ind, _ in enumerate(current_x) if not 0. == _])
omega_x = gamma_x | supp_x
print('omega_x', len(omega_x))
# print(sorted(list(omega_x)))
# print(gradient[sorted(list(gamma_x))])
# print(gradient[sorted(list(supp_x))])
bx = proj_argmax(current_x,
omega_x,
weight,
A,
lmbd,
max_iter=2000,
learning_rate=0.01)
re_tail = tail_proj(edges=edges,
weights=edge_weights,
x=bx,
g=1,
s=sparsity,
budget=sparsity - 1.,
nu=2.5,
max_iter=100,
err_tol=1e-8,
root=-1,
pruning='strong',
verbose=0)
re_nodes, _, _ = re_tail
psi_x = set(re_nodes)
prev_x = current_x
current_x = np.zeros_like(current_x)
current_x[list(psi_x)] = bx[list(psi_x)]
print('psi_x', len(np.nonzero(current_x)[0]))
diff_norm_x = np.linalg.norm(current_x - prev_x)
func_val, xtw, dense_term = proj_get_func_val(current_x, weight, A,
lmbd)
print(
'iter {}, func val: {:.5f}, iter_time: {:.5f}, diff_norm: {:.5f}'.
format(i, func_val,
time.time() - iter_time, diff_norm_x))
subgraph = set(np.nonzero(current_x)[0])
print('subgraph density', nx.density(nx.subgraph(graph, subgraph)))
if diff_norm_x <= epsilon:
break
run_time = time.time() - start_time
func_val, xtw, dense_term = proj_get_func_val(current_x, weight, A, lmbd)
print('final function value: {:.5f}'.format(func_val))
print('run time of whole algorithm: {:.5f}'.format(run_time))
subgraph = set(np.nonzero(current_x)[0])
return subgraph
cliques = nx.find_cliques(G) # Count and print the number of maximal cliques in G print(len(list(cliques))) ############################## Task 2 (Finding cliques (II)) # Import necessary modules import networkx as nx from nxviz import CircosPlot import matplotlib.pyplot as plt # Find the author(s) that are part of the largest maximal clique: largest_clique largest_clique = sorted(nx.find_cliques(G), key=lambda x: len(x))[-1] # Create the subgraph of the largest_clique: G_lc G_lc = nx.subgraph(G, largest_clique) # Create the CircosPlot object: c c = CircosPlot(G_lc) # Draw the CircosPlot to the screen c.draw() plt.show() ############################## Final Task ########################### # Find important users : nx.degree_centrality() # Find largest communities of collaborators : max_cliques() # recommmendation system : open triangle ############################## Task 1 (Finding important collaborators) # Compute the degree centralities of G: deg_cent
def inducer(graph, node):
nebs = nx.neighbors(graph, node)
sub_nodes = nebs + [node]
sub_g = nx.subgraph(graph, sub_nodes)
out_counts = np.sum(map(lambda x: len(nx.neighbors(graph,x)), sub_nodes))
return sub_g, out_counts, nebs
def optimize(instance,
sparsity,
threshold,
trade_off,
learning_rate,
max_iter,
epsilon=1e-3,
logger=None):
first_graph = instance['first_graph']
second_graph = instance['second_graph']
true_subgraph = instance['true_subgraph']
features = instance['weight']
A = adj_matrix(
second_graph
) # get adjacent matrix of second graph, used for density projection
first_graph_edges = np.array(first_graph.edges)
first_graph_edge_weights = np.ones(
first_graph.number_of_edges()) # edge weight, default 1
print('number of nodes in first graph', first_graph.number_of_nodes())
print('number of nodes in second graph', second_graph.number_of_nodes())
if first_graph.number_of_nodes() != second_graph.number_of_nodes():
raise ('error, wrong dual network input !!!')
num_nodes = first_graph.number_of_nodes()
num_edges_first_graph = first_graph.number_of_edges()
num_edges_second_graph = second_graph.number_of_edges()
if logger:
# print some basic information
logger.debug('-' * 5 + ' related info ' + '-' * 5)
logger.debug('algorithm: graph block-structured GHTP')
logger.debug('sparsity: {:d}'.format(sparsity))
logger.debug('max iteration: {:d}'.format(max_iter))
logger.debug('number of nodes: {:d}'.format(num_nodes))
logger.debug('number of edges in first graph: {:d}'.format(
num_edges_first_graph))
logger.debug('number of edges in second graph: {:d}'.format(
num_edges_second_graph))
logger.debug('density of first graph: {:.5f}'.format(
nx.density(first_graph)))
logger.debug('density of second graph: {:.5f}'.format(
nx.density(second_graph)))
logger.debug('density of true subgraph in second graph: {:.5f}'.format(
nx.density(nx.subgraph(second_graph, true_subgraph))))
logger.debug('-' * 5 + ' start iterating ' + '-' * 5)
start_time = time.time()
acc_proj_time = 0.
func = DualEMS(features, trade_off)
if logger:
print(sorted(true_subgraph))
true_x = np.zeros(num_nodes)
# print(type(true_subgraph))
true_x[list(true_subgraph)] = 1.
true_x = np.array(true_x)
true_obj_val, x_ems_val, y_ems_val, penalty = func.get_obj_val(
true_x, true_x)
print('ground truth values: {}, {}, {}, {}'.format(
true_obj_val, x_ems_val, y_ems_val, penalty))
current_x, current_y = func.get_init_x_zeros(
) # are there some other better initialization methods?
current_x += 1e-6 # start from zero, plus 1e-6 avoid divide by zero error
current_y += 1e-6
print('iteration start funval', func.get_obj_val(current_x, current_y))
for iter in range(max_iter): # external iteration
if logger:
logger.debug('iteration: {:d}'.format(iter))
prev_x, prev_y = np.copy(current_x), np.copy(
current_y) # store previous vectors for early termination
# handle first graph
grad_x = func.get_gradient(current_x, current_y)
iter_proj_time = 0.
if iter == 0: # from all zero vector
norm_grad_x = normalize_gradient(np.zeros_like(current_x), grad_x)
else:
norm_grad_x = normalize_gradient(current_x, grad_x)
start_proj_time = time.time()
# head projection for the connected constraint, so projection should be on first graph
re_head = head_proj(edges=first_graph_edges,
weights=first_graph_edge_weights,
x=norm_grad_x,
g=1,
s=sparsity,
budget=sparsity - 1.,
delta=1. / 169.,
max_iter=100,
err_tol=1e-8,
root=-1,
pruning='strong',
epsilon=1e-10,
verbose=0)
re_nodes, _, _ = re_head
iter_proj_time += (time.time() - start_proj_time)
print('head projection time for x: {:.5f}'.format(time.time() -
start_proj_time))
gamma_x = set(re_nodes)
indicator_x = np.zeros(num_nodes)
indicator_x[list(gamma_x)] = 1.
# there is no differene between using grad_x and norm_grad_x, because indicator_x is got from norm_grad_x
if iter == 0:
tmp_x = np.zeros_like(
current_x
) + learning_rate * norm_grad_x * indicator_x # start from all zeros
else:
tmp_x = current_x + learning_rate * norm_grad_x * indicator_x
omega_x = set([ind for ind, _ in enumerate(tmp_x) if not 0. == _])
# head projection for y
grad_y = func.get_gradient(current_y,
current_x) # note, reverse order x & y
# note, test not normalize
if iter == 0:
norm_grad_y = normalize_gradient(
np.zeros_like(current_y),
grad_y) # # note, is it necessary for density projection?
else:
norm_grad_y = normalize_gradient(current_y, grad_y)
# norm_grad_y = grad_y # note !!!
# note, should be positive for gradient, input for density projection should be positive
# note, why baojian's code does not consider positive value, head projection
abs_norm_grad_y = np.absolute(
norm_grad_y
) # take absolute value of gradient, since larger absolute value represent larger affection to objective function
np.set_printoptions(linewidth=3000)
# print(norm_grad_y)
# lmbd_list = [0.01, 0.05, 0.07, 0.08, 0.09, 0.1, 0.12, 0.15, 0.17, 0.2, 0.2, 0.2, 0.2, 0.21, 0.22, 0.23, 0.18, 0.18, 0.18, 0.17] # normalize
lmbd_list = [0.23] # normalize
# lmbd_list = [0.0001, 0.0002, 0.0003, 0.0004, 0.0005, 0.0006] # normalize
# sparsity_list = [250, 250, 260, 270, 270, 275, 275, 280, 280, 280, 265, 270, 275, 275, 280, 285, 260, 255, 250, 245] # normalize
sparsity_list = [275] # normalize
# sparsity_list = [50, 50, 50, 50, 55] # normalize
lmbd_sparsity_list = zip(lmbd_list, sparsity_list)
# sparsity_list = [50]
print('start head projection for y')
start_proj_time = time.time()
# gamma_y = density_projection(second_graph, norm_grad_y, A, threshold, min_sparsity, max_sparsity, step_sparsity, normalize=False)
gamma_y = density_projection(
second_graph,
abs_norm_grad_y,
A,
threshold,
lmbd_sparsity_list,
normalize=True,
true_subgraph=true_subgraph
) # test not normalize, need new lambda sparsity list
# gamma_y = density_projection(second_graph, abs_norm_grad_y, A, threshold, lmbd_sparsity_list, normalize=False, true_subgraph=true_subgraph) # test not normalize, need new lambda sparsity list
iter_proj_time += (time.time() - start_proj_time)
print('head projection time for y: {:.5f}'.format(time.time() -
start_proj_time))
#
#
# indicator_y = np.zeros(num_nodes)
# indicator_y[list(gamma_y)] = 1.
# if iter == 0:
# # tmp_y = np.zeros_like(current_y) + learning_rate * grad_y * indicator_y
# tmp_y = np.zeros_like(current_y) + learning_rate * norm_grad_y * indicator_y # todo, pls note that update gradient should be normalized gradient
# else:
# # tmp_y = current_y + learning_rate * grad_y * indicator_y
# tmp_y = current_y + learning_rate * norm_grad_y * indicator_y
#
# omega_y = set([ind for ind, _ in enumerate(tmp_y) if not 0. == _])
#
# print('omega_x', len(omega_x))
# print(sorted(list(omega_x)))
#
# print('omega_y', len(omega_y))
# print(sorted(list(omega_y)))
#
# print('intersect', len(omega_y & omega_x))
# print(sorted(list(omega_y & omega_x)))
#
# # break
#
# print('solve argmax')
# start_max_time = time.time()
# bx, by = func.argmax_obj_with_proj(current_x, current_y, omega_x, omega_y)
# print('solve argmax time {:.5f}'.format(time.time() - start_max_time))
#
# # break
#
# start_proj_time = time.time()
# # tail projection on first graph
# re_tail = tail_proj(edges=first_graph_edges, weights=first_graph_edge_weights, x=bx, g=1, s=sparsity, budget=sparsity - 1., nu=2.5, max_iter=100, err_tol=1e-8, root=-1, pruning='strong', verbose=0) # tail projection
# re_nodes, _, _ = re_tail
# iter_proj_time += time.time() - start_proj_time
# print('tail projection time for x: {:.5f}'.format(time.time() - start_proj_time))
# psi_x = set(re_nodes)
#
# current_x = np.zeros_like(current_x)
# current_x[list(psi_x)] = bx[list(psi_x)]
# current_x = normalize(current_x)
#
# lmbd_list = [0.01, 0.05, 0.07, 0.08, 0.09, 0.1]
# # lmbd_list = [0.006, 0.08]
# sparsity_list = [250, 250, 270, 270, 275, 275]
# lmbd_sparsity_list = zip(lmbd_list, sparsity_list)
#
# start_proj_time = time.time()
# # psi_y = density_projection(second_graph, by, threshold, min_sparsity, max_sparsity, step_sparsity, normalize=False)
# psi_y = density_projection(second_graph, by, A, threshold, lmbd_sparsity_list, normalize=False, true_subgraph=true_subgraph) # not normalize, not absolute value, since by is in [0, 1]
# iter_proj_time += (time.time() - start_proj_time)
#
# print('tail projetion time for y: {:.5f}'.format(time.time() - start_proj_time))
#
# current_y = np.zeros_like(current_y)
# print('1', len(np.nonzero(by)[0]))
# print('by nonzero', sorted(list(np.nonzero(by)[0])))
# print('1v', len(np.nonzero(bx)[0]))
# print('2', len(psi_y))
# print('psi_y', sorted(list(psi_y)))
# print('2v', len(psi_x))
# current_y[list(psi_y)] = by[list(psi_y)]
# print('3', len(np.nonzero(current_y)[0]))
# print('3v', len(np.nonzero(current_x)[0]))
# current_y = normalize(current_y)
# print('4', len(np.nonzero(current_y)[0]))
# print('4v', len(np.nonzero(current_x)[0]))
#
# print('{} iteration funval'.format(iter), func.get_obj_val(current_x, current_y))
#
# acc_proj_time += iter_proj_time
#
# if logger:
# print('iter proj time: {:.5f}'.format(iter_proj_time))
#
# diff_norm = np.sqrt(np.linalg.norm(current_x - prev_x) ** 2 + np.linalg.norm(current_y - prev_y) ** 2)
# if logger:
# logger.debug('difference norm: {}'.format(diff_norm))
#
# if diff_norm < epsilon:
# break
#
run_time = time.time() - start_time
if logger:
pass
return current_x, current_y, run_time
def connected_components(graph, nbunch):
g = nx.subgraph(graph, nbunch)
ccomponents = nx.connected_components(g)
return ccomponents
def draw_graph(self, ax=None, node_size=8):
""" Draws the state graph of the maze in a fancy way. """
if ax is None:
fig, ax = plt.subplots(1)
graph = self.compute_graph()
d = nx.shortest_path(graph, graph.start)
p_goal = d[graph.goal]
d_goal = len(p_goal)
subgraph = nx.subgraph(graph, d.keys())
n_levels = max(map(len, d.values()))
nodespos = dict()
max_nodes_in_level = 0
nodes = [graph.start]
nodespos[graph.start] = [-0.5, 0]
for i in range(1, n_levels + 2):
nodes = [[
n for n in nx.DiGraph.successors(graph, m)
if len(d[n]) == i + 1
] for m in nodes]
nodes = sum(nodes, [])
seen = set()
seen_add = seen.add
nodes = [n for n in nodes if n not in seen and not seen_add(n)]
#nodes = list(set(sum(nodes, [])))
# place the nodes of the solution on the left
l = [n for n in nodes if n in p_goal]
if l != []:
nodes.remove(l[0])
nodes = [l[0]] + nodes
if len(nodes) > max_nodes_in_level:
max_nodes_in_level = len(nodes)
for j, n in enumerate(nodes):
nodespos[n] = [j - 1.0 * len(nodes) / 2, -i]
if ax is None:
fig, ax = subplots(1)
main_edges = [(n1, n2) for (n1, n2) in subgraph.edges()
if (len(d[n2]) - len(d[n1]) == 1)]
other_edges = [e for e in subgraph.edges() if e not in main_edges]
nx.draw_networkx_edges(nx.Graph(subgraph),
edgelist=other_edges,
ax=ax,
pos=nodespos,
width=0.5,
alpha=.4)
nx.draw_networkx_edges(nx.Graph(subgraph),
edgelist=main_edges,
ax=ax,
pos=nodespos,
width=2)
nx.draw_networkx_nodes(subgraph,
pos=nodespos,
ax=ax,
node_size=node_size)
for n, col in ([graph.start, 'g'], [graph.goal, 'b']):
nx.draw_networkx_nodes(graph,
nodespos, [n],
ax=ax,
node_size=1.5 * node_size,
node_color=col)
ax.set_ylim(-n_levels, 1)
ax.set_xlim(-1.0 * max_nodes_in_level / 2 - 1,
1.0 * max_nodes_in_level / 2)
ax.set_axis_off()
def compute_score(self):
"""
Computes a score for the maze.
This is an example of scoring function which implements a
few of my favorite criteria for a good maze:
- The solution is unique and long.
- There are plenty of loops and bagends.
- Many openings, many false endings which make the maze
difficult to solve backwards.
This is all very subjective and incomplete. To grow mazes
according to other criteria, overwrite this function in a
subclass of Vmaze.
"""
score = 1.0
graph = self.compute_graph()
if len(graph.nodes()) == 0:
return 0
shortest = nx.shortest_path(graph, graph.start)
# Check that there is only one solution.
if graph.goal not in shortest.keys():
return 0
# Check that there is one solution exactly
gen = nx.all_simple_paths(graph, graph.start, graph.goal)
next(gen)
try:
next(gen)
except StopIteration:
pass
else:
return 1.0
distances = {n: len(p) for n, p in shortest.items()}
path_goal = shortest[graph.goal]
d_goal = distances[graph.goal]
subgraph = nx.subgraph(graph, shortest.keys())
# Reward the length of the minimal solution
score *= d_goal**2
# Reward the openings, and the states that are nearer
# than the solution in general
openings_1 = len([d for n, d in distances.items() if d <= 2])
openings_2 = len([d for n, d in distances.items() if d <= d_goal])
score *= (openings_1 * openings_2**2)
# Reward the "loops"
n_loops = len([
e for e in subgraph.edges()
if abs(distances[e[0]] - distances[e[1]]) > 1
])
score *= (n_loops**0.5)
# Reward the false endings
paths_to_end = nx.shortest_path(graph, target=graph.goal)
n_endings = len(
[path for path in paths_to_end.values() if len(path) < 3])
score *= n_endings
return score
def simulated_annealing(graph,
num_buses,
size_bus,
constraints,
test_mode=False):
all_names = graph.nodes()
x = all_names[0]
print(graph[x])
print(graph[x]['weight'])
CD = constraint_dictionary(all_names, constraints)
#assignment = random_assignment(all_names, num_buses, size_bus)
assignment = greedy(graph, num_buses, size_bus, constraints)
groups = assignment_to_groups(assignment)
assert is_valid(graph, num_buses, size_bus, assignment)
T0 = 1.0
iterations = 5000
student_status = calculate_student_status(graph, constraints, assignment)
starting_score = evaluate_assignment(graph, constraints, assignment)
current_score = starting_score
for i in range(iterations):
T = T0 * (1 - i / iterations)
swap = propose_swap(assignment, num_buses, size_bus, groups=groups)
if swap is None:
continue
val, student_status_swap = swap_value(graph, constraints, CD,
student_status, assignment, swap)
if val >= 0 or random.random() < np.exp(val / T):
group_A, A, group_B, B = swap
if A is not None:
assignment[A] = group_B
groups[group_B].append(A)
groups[group_A].remove(A)
assignment[B] = group_A
groups[group_A].append(B)
groups[group_B].remove(B)
student_status = student_status_swap
if (test_mode):
new_score = evaluate_assignment(graph, constraints, assignment)
student_status_check = calculate_student_status(
graph, constraints, assignment)
for s in student_status_check:
assert student_status[s] == student_status_check[s]
if new_score != current_score + val:
print(new_score, current_score, val)
print(swap)
print(student_status)
print(assignment_to_groups(assignment))
print(constraints)
nodes = []
for x in assignment:
if assignment[x] == group_A or assignment[x] == group_B:
nodes.append(x)
import matplotlib.pyplot as plt
subgraph = nx.subgraph(graph, nodes)
nx.draw(subgraph, with_labels=True, font_weight='bold')
plt.show()
assert False
current_score = new_score
assert is_valid(graph, num_buses, size_bus, assignment)
#print('Score:', evaluate_assignment(graph, constraints, assignment))
return assignment
def _build_list(self):
print "using disconnectivity analysis to find minima to untrap"
self.minpairs = deque()
graph = TSGraph(self.database).graph
cclist = list(nx.connected_components(graph))
# get the largest cluster
group1 = cclist[0]
min1 = sorted(group1, key=lambda m: m.energy)[0]
if not min1 == self.database.minima()[0]:
# make sure that the global minimum is in group1
print "warning, the global minimum is not the in the largest cluster."
# compute the energy barriers for all minima in the cluster
subgraph = nx.subgraph(graph, group1)
energy_barriers = self._compute_barriers(subgraph, min1)
# sort the minima by the barrier height divided by the energy difference
weights = [(m, np.abs(barrier) / np.abs(m.energy - min1.energy))
for (m, barrier) in energy_barriers.iteritems()]
weights.sort(key=lambda v: 1. / v[1])
self.minpairs = deque()
for min2, w in weights:
if len(self.minpairs) > self.list_len:
break
if not self.is_good_pair(min1, min2):
continue
self.minpairs.append((min1, min2))
if True:
# print some stuff
print " untrap analysis: minimum", min2._id, "with energy", min2.energy, "barrier", energy_barriers[min2], "untrap weight", w
