def planning(env, path_nodes, start, goal):
"""
env : environment parameter
path_node : list of [x, y] represents the path nodes
"""
# create the nodes
G = nx.graph()
G.add_node(0, loc = start)
for i in len(path_nodes):
G.add_node(i+1, loc = path_nodes[i])
G.add_node(len(path_nodes) + 1, loc = path_nodes[i])
# create the edges
for i in range(len(path_nodes) + 2):
for j in range(len(path_nodes) + 2):
if collision_detection(env = env, graph = G , node1 = i, node2 = j, step = 0.1):
G.add_edge(i, j)
# use bellman_ford algorithm to compute the distance
return distance, node_list
def get_networkx_graph(self):
"""
将网络拓扑转化为图的形式表示
:return:
"""
graph = nx.graph()
return graph
def get_future_link_graph(self):
Graph = nx.graph()
for line in self.future_dataset.readlines():
line = line.strip()
col = line.split(';')
node1_id = int(col[0])
node2_id = "p%s" %(int(col[1]))
Graph.add_edge(node1_id, node2_id)
return Graph
def create_graph(data):
G = nx.graph() #empty graph
for i, elrow in data.iterrows():
g.add_edge(elrow[0], elrow[1], attr_dict=elrow[2:].to_dict())
if __name__ == "__main__":
a = import_data('https://gist.githubusercontent.com/brooksandrew/e570c38bcc72a8d102422f2af836513b/raw/89c76b2563dbc0e88384719a35cba0dfc04cd522/edgelist_sleeping_giant.csv')
b = import_data('https://gist.githubusercontent.com/brooksandrew/f989e10af17fb4c85b11409fea47895b/raw/a3a8da0fa5b094f1ca9d82e1642b384889ae16e8/nodelist_sleeping_giant.csv')
create_graph(a)
def graph(next_ids):
g = nx.graph()
g.add_nodes_from(next_ids) #creates nodes from the users ids
for j in next_ids.items():
g.add_edges_from([ (j,k) for k in y ]) #adds edges to the id nodes
data = json_graph.node_link_data(G)
with open('data.json', 'w', encoding ='utf-8') as f: json.dump(data, f, ensure_ascii=False, indent=4)
diameter = nx.diameter(n_graph) #calculates diameter
avg_distance = nx.average_shortest_path_length(n_graph) #calculates the average distance
print("Diameter: ", diameter, "\nAverage Distance: ", avg_distance)
def pdist(x, dist=distance, threshold=7):
#D = np.zeros(len(x), len(x))
G = nx.graph()
for i in range(len(x)):
for j in range(i+1, len(x)):
d = dist(x[i], x[j])
if d<threshold:
G.add_edge(i,j,weight)
return G
def load_graph(graph_name):
if graph_name=='':
return
if graph_name == 'Board Ex':
## Board example - Toy Graph
G = nx.Graph()
G.add_edge(1,2)
G.add_edge(1,5)
G.add_edge(2,3)
G.add_edge(2,4)
G.add_edge(3,4)
G.add_edge(3,5)
G.add_edge(4,6)
G.add_edge(5,6)
# G.add_edge(5,7)
# G.add_edge(6,7)
#print '--- Board Example ---'
elif graph_name == 'Karate Club':
G = nx.karate_club_graph()
elif graph_name == "Erdos-Renyi":
n=100 # 10 nodes
p= 0.5# 20 edges
G=nx.gnp_random_graph(n,p) #binomial_graph which is also somtims called the Erdos-Renyi grap h.
elif graph_name == "NewWatStr SW":
k=2
p=0.5
G = nx.newman_watts_strogatz_graph(n, k, p)
elif graph_name == "Kronecker":
## TODO: implement Kronecker
G = nx.graph()
#
elif graph_name == "edgelist":
G = nx.read_edgelist('../demo_graphs/out.contact',comments="%",delimiter="\t")
G = nx.read_edgelist('../demo_graphs/netscience.txt')
#LCCG = sorted(nx.connected_components(G), key = len, reverse=True)
cc = sorted(list(nx.connected_component_subgraphs(G)), key = len, reverse=True)
G = cc[0]
else:
G = nx.graph()
return (graph_name,G)
def visualize(self, ajmatrix, ntype="bit"):
"""
ajmatrix -> adjacency matrix of graph(array like)
ntype -> default: bit(string) or add list of node(list)
"""
if ntype == "bit":
edge = [
format(i, "0%sb" % (self.n_qubit)) for i in range(self.nodes)
]
print(edge)
else:
edge = ntype
G = nx.graph()
def get_stations():
with open("tube.csv",mode="r",encoding ="utf-8") as stations:
reader = csv.reader(stations)
for line in reader:
print(line[0])
print(line[1])
print(line[2:])
print()
data = {"line name":line[0],"edge colour":line[1]}
print(data)
print()
print()
tube_graph = nx.graph()
tube_graph.add_path(line[2:],data)
def get_stations():
with open("tube.csv", mode="r", encoding="utf-8") as stations:
reader = csv.reader(stations)
for line in reader:
print(line[0])
print(line[1])
print(line[2:])
print()
data = {"line name": line[0], "edge colour": line[1]}
print(data)
print()
print()
tube_graph = nx.graph()
tube_graph.add_path(line[2:], data)
def report(antcompend): G=nx.graph() G=from_numpy_matrix(antcompend) noden=nx.number_of_nodes(G) for n in range(noden): for nn in range(noden): if n==nn: pass G.add_edge(n,nn,weigh=route[n][nn]) pos=nx.spring_layout(T) plt.figure(figsize=(8,8)) nx.draw_networkx_nodes(G,pos) nx.draw_networkx_edges(G,pos,) plt.show()
def load_graph(graph_name):
if graph_name == '':
return
if graph_name == 'BoardEx':
## Board example - Toy Graph
G = nx.Graph()
G.add_edge(1, 2)
G.add_edge(2, 3)
G.add_edge(2, 4)
G.add_edge(3, 4)
G.add_edge(3, 5)
G.add_edge(4, 6)
G.add_edge(5, 6)
G.add_edge(5, 7)
G.add_edge(6, 7)
#print '--- Board Example ---'
elif graph_name == 'KarateClub':
G = nx.karate_club_graph()
elif graph_name == "Erdos-Renyi":
n = 68 # 10 nodes
m = 68 # 20 edges
G = nx.gnm_random_graph(
n, m
) #binomial_graph which is also somtims called the Erdos-Renyi grap h.
elif graph_name == "NewWatStrSW":
k = 2
p = 0.5
G = nx.newman_watts_strogatz_graph(n, k, p)
elif graph_name == "Kronecker":
## TODO: implement Kronecker
G = nx.graph()
#
else:
G = nx.graph()
return (graph_name, G)
def report(antcompend):
G = nx.graph()
G = from_numpy_matrix(antcompend)
noden = nx.number_of_nodes(G)
for n in range(noden):
for nn in range(noden):
if n == nn:
pass
G.add_edge(n, nn, weigh=route[n][nn])
pos = nx.spring_layout(T)
plt.figure(figsize=(8, 8))
nx.draw_networkx_nodes(G, pos)
nx.draw_networkx_edges(
G,
pos,
)
plt.show()
def gen_independent_graph():
global ind_graph
global followers_dict
ind_graph=Ddict(dict)
big_chunk=read_timeline()
for chunk in big_chunk:
first=chunk[0].split(' ')
hashtag=first[0].strip()
time=first[1].strip()
ind_graph[hashtag][time]=nx.Graph()
chunk=chunk[1:]
for user in chunk:
user=user.split(' ')
time=user[0].strip()
id=user[1].strip()
if time in ind_graph[hashtag]:
graph=ind_graph[hashtag][time]
graph.add_node(user)
else:
graph=nx.graph()
graph.add_node(user)
ind_graph[hashtag][time]=graph
for hashtag in ind_graph:
for time in ind_graph[hashtag]:
graph=ind_graph[hashtag][time]
for node in graph.nodes():
for other in graph.nodes():
if node!=other:
if node in followers_dict[other]:
if other in followers_dict[node]:
graph.add_edge(node,other)
ind_graph[hashtag][time]=graph
gen_dependent_graph(big_chunk)
def gen_dependent_graph(big_chunk):
global dep_graph
dep_graph = {}
for chunk in big_chunk:
hashtag = chunk[0].split(' ').strip()
chunk = chunk[1:]
graph = nx.graph()
for user in chunk:
id = user.split(' ')[1]
graph.add_node(id)
dep_graph[hashtag] = graph
for hashtag in dep_graph:
graph = dep_graph[hashtag]
for node in graph.nodes():
for other in graph.nodes():
if node != other:
if node in followers_dict[other]:
if other in followers_dict[node]:
graph.add_edge(node, other)
dep_graph[hashtag] = graph
def gen_dependent_graph(big_chunk):
global dep_graph
dep_graph={}
for chunk in big_chunk:
hashtag=chunk[0].split(' ').strip()
chunk=chunk[1:]
graph=nx.graph()
for user in chunk:
id=user.split(' ')[1]
graph.add_node(id)
dep_graph[hashtag]=graph
for hashtag in dep_graph:
graph=dep_graph[hashtag]
for node in graph.nodes():
for other in graph.nodes():
if node!=other:
if node in followers_dict[other]:
if other in followers_dict[node]:
graph.add_edge(node,other)
dep_graph[hashtag]=graph
def gen_independent_graph():
global ind_graph
global followers_dict
ind_graph = Ddict(dict)
big_chunk = read_timeline()
for chunk in big_chunk:
first = chunk[0].split(' ')
hashtag = first[0].strip()
time = first[1].strip()
ind_graph[hashtag][time] = nx.Graph()
chunk = chunk[1:]
for user in chunk:
user = user.split(' ')
time = user[0].strip()
id = user[1].strip()
if time in ind_graph[hashtag]:
graph = ind_graph[hashtag][time]
graph.add_node(user)
else:
graph = nx.graph()
graph.add_node(user)
ind_graph[hashtag][time] = graph
for hashtag in ind_graph:
for time in ind_graph[hashtag]:
graph = ind_graph[hashtag][time]
for node in graph.nodes():
for other in graph.nodes():
if node != other:
if node in followers_dict[other]:
if other in followers_dict[node]:
graph.add_edge(node, other)
ind_graph[hashtag][time] = graph
gen_dependent_graph(big_chunk)
def __init__(self):
self.graph = nx.graph()
self.userList = []
def __init__(self):
self.G = networkx.graph()
import networkx as nx def dfs(graph, s): """ graph - networkx graph. Which graph would you like to search? s - where is your starting node? """ return None if __name__ == "__main__": g = nx.graph() g.add_edges_from()
best = 0 # best graph split
while True: # iterate until we use all the number of edges
gnStep(g)
q = gnModularity(g, orig, m)
if q > best: # if the modularity of this method of partitioning is better
best = q
bestcom = nx.connected_components(
g) # best way of partitioning the graph
if g.number_of_edges() == 0:
break # break the loop when we get to this part of the algorithm
if best > 0:
print("Modularity (Q): %f " % best)
print("Graph communities:", bestcom)
else:
print("Modularity (Q): %f" % best)
gr = sys.argv[1] # read input graph file
g = nx.graph() # initialize is using networkx
buildGraph(g, gr, ",") # build the graph
n = g.number_of_nodes()
a = nx.adj_matrix(g) # adjacency matrix
m = 0 # number of edges
for i in range(n):
for j in range(n):
m += a[i, j]
m = m / 2 # normalize
gnRun(g, Deg(a, g.nodes()),
m) # run the algorithm after getting the weighted degree for each node
def growgraph(usenx= True, force_connected = True, sparse = True, plot = False, directed = True, randomgrowth=False, wholegrowth=False,growthfactor=100, num_measurements = 10, verbose = True, connected = True, plotx = 'nodegrowth', ploty = 'maxclique', ploty2 = 'modularity',drawgraph = 'moral', draw= True):
# initialize database
graph_db = neo4j.GraphDatabaseService()
#get the pickled dictionary of nodes and node names that is made while the csv file is loaded.
try:
nodes = open('nodes.p','r')
nodes = pickle.load(nodes)
except:
print 'Your graph is empty. Please loab a graph into the database.'
1/0
data = []
if graph_db.size < 2:
print 'Your graph is empty. Please load a graph into the database.'
1/0
#make a list of all the nodes in the database to randomly choose from, but only if force connected graph option is false.
if force_connected == False:
possiblenodes = nodes
print 'Graph will not be fully connected, use "force_connected = True" if you want it to be fully connected'
# here we figure out at what points to measure if the user wants a spare measumement or to measure every time a node is added. This speeds up big graph growths a lot!
if sparse == True:
sparsemeasurements = [6,7,8,9,growthfactor]
measurements = np.logspace(1, (int(len(str(growthfactor))))-1, num_measurements)
for x in measurements:
sparsemeasurements.append(int(x))
else:
sparsemeasurements = range(2,growthfactor)
# this will actually only work for directed graph because we are moralizing, but I want to leave the option for later. Perhaps I can just skip moralization for undirected graph.
if directed:
graph = nx.DiGraph()
else:
graph = nx.graph()
# this will grow the whole graph. Takes a while for big ones!
if wholegrowth == True:
growthfactor = len(nodes)
# pick an initial node to start from
initial_node = random.choice(nodes.keys()) #randomly get a node to add from dictionary
if verbose:
print 'Starting from ' + initial_node
used = 0
#measure the nodegrowth
nodegrowth = graph.size()
while nodegrowth < growthfactor: # make sure we aren't above how many nodes we want to measure in the graph
if force_connected == True:
if nodegrowth > 1:
possiblenodes = graph.nodes() # get all nodes in graph.
fromnode = random.choice(possiblenodes) #pick random node in graph to grow from.
if verbose:
print 'Using ' + str(fromnode) + ' to find new node'
else: #this is because you can't do random from one node at the start.
fromnode = initial_node
if verbose:
print 'using initial node'
used = used+1
if used > 5:
1/0
fromnode = nodes[fromnode] #get DB version of the node.
#get all relationpships in graph DB for that node
new_node_rels = list(graph_db.match(end_node = fromnode, bidirectional=True))
new_rel = random.choice(new_node_rels)
# is the new node a parentt or child of the node we are growing from?
if new_rel.end_node == fromnode:
new_node = new_rel.start_node
if new_rel.start_node == fromnode:
new_node = new_rel.end_node
assert new_node != fromnode
if force_connected == False:
new_node = random.choice(possiblenodes.values())
print 'adding' + str(new_node)
# go through the list of edges that have the new node as a part of it, and only add the edge if they are between the new node and a node in the graph already.
rels = list(graph_db.match(start_node=new_node)) #query graph for edges from that node
for edge in rels:
newnodename = edge.start_node.get_properties()
newnodename = newnodename.get('name')
newnodename = newnodename.encode()
endnodename = edge.end_node.get_properties()
endnodename = endnodename.get('name')
endnodename = endnodename.encode()
if newnodename not in graph: #check to see if new node is in graph
graph.add_node(newnodename) # add if not
if verbose:
print 'added ' + str(newnodename)
if endnodename in graph: #check to see if end node is in graph
graph.add_edge(newnodename, endnodename) #add it if it is
if verbose:
print 'connected ' + newnodename +' to '+ endnodename
rels = list(graph_db.match(end_node=new_node)) #query graph for edges to that node
for edge in rels:
newnodename = edge.end_node.get_properties()
newnodename = newnodename.get('name')
newnodename = newnodename.encode()
startnodename = edge.start_node.get_properties()
startnodename = startnodename.get('name')
startnodename = startnodename.encode()
if newnodename not in graph: #check to see if new node is in graph
graph.add_node(newnodename) # add if not
if verbose:
print 'added ' + str(newnodename)
if startnodename in graph: #check to see if end node is in graph
graph.add_edge(startnodename, newnodename) #add it if it is
if verbose:
print 'connected ' + startnodename +' to '+ newnodename
nodegrowth = len(graph.nodes())
if verbose:
print 'Graph has ' + str(nodegrowth) + ' nodes.'
edgegrowth = len(graph.edges())
if verbose:
print 'Graph has ' + str(edgegrowth) + ' edges.'
if nodegrowth in sparsemeasurements:
start_time = time()
try:
modgraph = nx.Graph(graph) #make it into a networkx graph
partition = community.best_partition(modgraph) #find the partition with maximal modularity
modularity = community.modularity(partition,modgraph) #calculate modularity with that partition
except:
modularity = 0.0
moral = nx.to_numpy_matrix(graph)
if usenx == True:
moralized = pybayes.moralize(graph)
moralized, moral_edges = octave.moralize(moral)
ismoral = False
tries = 0
while ismoral == False and tries < 5: #sometimes oct2py takes too long to return I think
try:
moralized, moral_edges = octave.moralize(moral)
ismoral = True
except:
ismoral = False
print 'Octave Error, trying again!'
tries = tries + 1
order = range(len(moralized)+1)
order = order[1:]
triangulated, cliques, fill_ins = octave.triangulate(moralized,order)
istriangulated = False
tries = 0
while istriangulated == False and tries < 5: #sometimes oct2py takes too long to return I think
try:
triangulated, cliques, fill_ins = octave.triangulate(moralized,order)
istriangulated = True
tries = 5
except:
istriangulated = False
print 'Octave Error, trying again!'
tries = tries + 1
cliquesizes = []
for x in cliques:
size = len(x)
cliquesizes.append(size)
maxclique = max(cliquesizes)
avgclique = np.mean(cliquesizes)
end_time = time()
run_time = end_time - start_time
data.append([nodegrowth, len(graph.edges()), modularity, maxclique,avgclique,run_time])
sparsemeasurements.remove(nodegrowth) #so we don't calculate clique size more than once!
if verbose:
print 'took ' + str(run_time) + ' to run last computation'
print 'Growing graph with ' + str(nodegrowth) + ' nodes, ' + str(modularity) + ' modularity, max clique of ' + str(maxclique) + ', ' + str(len(graph.edges())) + ' edges.'
if plot == True:
df = pd.DataFrame(data, columns= ('nodegrowth','edgegrowth', 'modularity','maxclique','avgclique','run_time'))
plt.close()
fig = plt.figure(figsize=(24,16))
ax = fig.add_subplot(1,1,1)
ax2 = ax.twinx()
y1 = df[ploty]
y2 = df[ploty2]
x = df[plotx]
ax.set_xlabel('%s in Graph' %(plotx),fontsize=20)
line1, = ax.plot(x, y1, label = ploty)
line2, = ax2.plot(x, y2, label = ploty2, color = 'red')
ax.set_ylabel(ploty,fontsize=20)
ax2.set_ylabel(ploty2, fontsize=20)
plt.suptitle('Graph Growth', fontsize=30)
plt.legend((line1,line2),(ploty , ploty2), loc='upper center', frameon=False, fontsize=20)
plt.show()
if draw == True:
if drawgraph == 'triangulated':
G = nx.from_numpy_matrix(triangulated)
elif drawgraph == 'moral':
G = nx.from_numpy_matrix(moralized)
elif drawgraph == 'directed':
G = graph()
# position is stored as node attribute data for random_geometric_graph
plt.close()
pos = nx.random_layout(G)
# find node near center (0.5,0.5)
dmin=1
ncenter=0
for n in pos:
x,y=pos[n]
d=(x-0.5)**2+(y-0.5)**2
if d<dmin:
ncenter=n
dmin=d
# color by path length from node near center
p=nx.single_source_shortest_path_length(G,ncenter)
plt.figure(figsize=(15,15))
nx.draw_networkx_edges(G,pos,nodelist=[ncenter],alpha=0.2)
nx.draw_networkx_nodes(G,pos,nodelist=p.keys(),
node_size=20,
node_color=p.values(),
cmap=plt.cm.Reds_r)
plt.xlim(-0.05,1.05)
plt.ylim(-0.05,1.05)
plt.axis('off')
plt.show()
df = pd.DataFrame(data, columns= ('nodegrowth','edgegrowth', 'modularity','maxclique','avgclique','run_time'))
return(df)
import networkx as nx
import random
edge_list = [(1, 2), (1, 4), (1, 5), (2, 4), (2, 6), (3, 5), (3, 6), (4, 5),
(5, 6), (7, 8), (7, 10), (8, 9), (9, 10)]
# create G2 and Gc using edge_list
n_edges_added_for_weak_connectedness = 0
n_edges_added_for_strong_connectedness = 0
G2 = nx.graph()
G2.add_node(range(1, 11))
G2.add_edge(x for x in edge_list)
while (nx.is_strongly_connected(G2) == False):
u = random.choice(list(G2.nodes()))
v = random.choice(list(G2.nodes()))
if (u != v):
if (nx.is_weakly_connected(G2) == False):
G2.add_edge(u, v)
n_edges_added_for_strong_connectedness += 1
n_edges_added_for_weak_connectedness += 1
else:
G2.add_edge(u, v)
n_edges_added_for_strong_connectedness += 1
Gc = G2.copy()
def __init__(self, populace, home_infectivity, school_infectivity,
work_infectivity):
self.graph = nx.graph()
self.home_infectivity = home_infectivity
self.school_infectivity = school_infectivity
self.work_infectivity = work_infectivity
def grow_graph(reverserandom = False, outgoingrandom = False, incomingrandom = False, totalrandom = False, usenx= True, force_connected = True, sparse = True, plot = False, directed = True, randomgrowth=False, wholegrowth=False,growthfactor=100, num_measurements = 10, verbose = True, connected = True, plotx = 'nodegrowth', ploty = 'maxclique', ploty2 = 'modval',drawgraph = 'triangulated', draw= True, drawspectral = True, getgraph = True):
random = False #set as false, gets made to truth later if the person passes a type of random graph they want
#make sure user does not want to draw and plot at the same time.
if plot == True:
draw = False
if draw == True:
plot = False
#you cannot plot and draw at the same time
# initialize database
graph_db = database()
#get the pickled dictionary of nodes and node names (for the database) that is made while the csv file is loaded.
try:
nodes = open('nodes.p','r')
nodes = pickle.load(nodes)
except:
print 'Your graph is empty. Please load a graph into the database.'
1/0
data = []
if graph_db.size < 2:
print 'Your graph is empty. Please load a graph into the database.'
1/0
# this will grow the whole graph. Takes a while for big ones!
if wholegrowth == True:
growthfactor = len(nodes)
#make a list of all the nodes in the database to randomly choose from, but only if force connected graph option is false.
if force_connected == False:
possiblenodes = nodes
print 'Graph will not be fully connected, use "force_connected = True" if you want it to be fully connected'
# here we figure out at what points to measure if the user wants a spare measumement or to measure every time a node is added. This speeds up big graph growths a lot!
if sparse == True:
sparsemeasurements = [10,15,growthfactor]
measurements = np.logspace(1, (int(len(str(growthfactor))))-1, num_measurements)
for x in measurements:
sparsemeasurements.append(int(x))
else:
sparsemeasurements = range(2,growthfactor)
# this will actually only work for directed graph because we are moralizing, but I want to leave the option for later. Perhaps I can just skip moralization for undirected graphs.
if directed:
graph = nx.DiGraph()
else:
graph = nx.graph()
# pick an initial node to start from
initial_node = np.random.choice(nodes.keys()) #randomly get a node to add from dictionary
if verbose:
print 'Starting from ' + initial_node
initial_used = 0 #keep track of how many times the initial node was used
nodegrowth = graph.number_of_nodes()
edgegrowth = graph.number_of_edges()
while nodegrowth < growthfactor: # make sure we aren't above how many nodes we want to measure in the graph
#start off finding a node to add to the graph.
if force_connected == True:
try:
possiblenodes = graph.nodes() # get all nodes in graph.
fromnode = Random.choice(possiblenodes) #pick random node in graph to grow from(add one of its nieghbors). This uses the random module, not np.random
if verbose:
print 'Using ' + str(fromnode) + ' to find new node'
except: #this is because you can't do random from one node at the start.
fromnode = initial_node
graph.add_node(fromnode)
if verbose:
print 'using initial node'
initial_used = initial_used+1
if initial_used > 5:
1/0
fromnode = nodes[fromnode] #get DB version of the node.
#get all relationships in graph DB for that node so we can pick
new_node_rels = list(graph_db.match(end_node = fromnode, bidirectional=True))
new_rel = Random.choice(new_node_rels) #randomly pick one of them, thus picking a node to add to graph. This uses the random module, not np.random
# is the new node a parent or child of the node we are growing from?
if new_rel.end_node == fromnode:
new_node = new_rel.start_node
if new_rel.start_node == fromnode:
new_node = new_rel.end_node
assert new_node != fromnode
if force_connected == False: # if not connected, we can just pick from the pickled dictionary of nodes in the database
new_node = np.random.choice(possiblenodes.values())
print 'adding' + str(new_node)
#add the nodes to the graph, connecting it to nodes in the graph that it is connected to.
# go through the list of edges that have the new node as a part of it, and only add the edge if they are between the new node and a node in the graph already.
rels = list(graph_db.match(start_node=new_node)) #query graph for edges from that node
for edge in rels:
#get the string name of the node
newnodename = edge.start_node.get_properties()
newnodename = newnodename.get('name')
newnodename = newnodename.encode()
endnodename = edge.end_node.get_properties()
endnodename = endnodename.get('name')
endnodename = endnodename.encode()
if newnodename not in graph: #check to see if new node is in graph
graph.add_node(newnodename) # add if not
if verbose:
print 'added ' + str(newnodename)
if endnodename in graph: #check to see if end node is in graph
graph.add_edge(newnodename, endnodename) #add it if it is
if verbose:
print 'connected ' + newnodename +' to '+ endnodename
rels = list(graph_db.match(end_node=new_node)) #query graph for edges to that node
for edge in rels:
newnodename = edge.end_node.get_properties()
newnodename = newnodename.get('name')
newnodename = newnodename.encode()
startnodename = edge.start_node.get_properties()
startnodename = startnodename.get('name')
startnodename = startnodename.encode()
if newnodename not in graph: #check to see if new node is in graph
graph.add_node(newnodename) # add if not
if verbose:
print 'added ' + str(newnodename)
if startnodename in graph: #check to see if end node is in graph
graph.add_edge(startnodename, newnodename) #add it if it is
if verbose:
print 'connected ' + startnodename +' to '+ newnodename
#measure the nodegrowth and edge growth
nodegrowth = len(graph.nodes())
edgegrowth = len(graph.edges())
if verbose:
print 'Graph has ' + str(nodegrowth) + ' nodes.'
if verbose:
print 'Graph has ' + str(edgegrowth) + ' edges.'
#here I make a few random graphs to compare the growth of the real graph. They are all the same size and same edges, but different controls.
# Reverse the direction of the edges. The in-degree distribution in this graph is power-law, but the out-degree is exponential tailed. So this is just a check that degree distribution is irrelevant.
if reverserandom == True:
random_graph = graph.reverse()
random = True
# Keep the number of outgoing links for each node the same, but randomly allocating their destinations. This should break modularity, put preserves out-degree.
if outgoingrandom == True:
random_graph = graph.copy()
for edge in random_graph.edges():
parent = edge[0]
child = edge[1]
random_graph.remove_edge(parent,child)
newchild = parent
while newchild == parent: #so that we don't get a self loop.
newchild = np.random.choice(graph.nodes())
random_graph.add_edge(parent,newchild)
random = True
# Same thing, but this time fixing the number of incoming links and randomly allocating their origins. Likewise, but preserves in-degree.
if incomingrandom ==True:
random_graph = graph.copy()
for edge in random_graph.edges():
parent = edge[0]
child = edge[1]
random_graph.remove_edge(parent,child)
newparent = child
while newparent == child:
newparent = np.random.choice(graph.nodes())
random_graph.add_edge(newparent,child)
random = True
#gives a graph picked randomly out of the set of all graphs with n nodes and m edges. This preserves overall degree, and number of nodes/edges, but is completeley random to outdegree/indegree.
if totalrandom == True:
numrandomedges = graph.number_of_edges()
numrandomnodes = graph.number_of_nodes()
random_graph = nx.dense_gnm_random_graph(numrandomnodes, numrandomedges)
random_graph = random_graph.to_directed()
random = True
#here is where we measure everything about the graph
if nodegrowth in sparsemeasurements: #we only measure every now and then in sparse.
start_time = time()
if nodegrowth > 5:
modgraph = nx.Graph(graph) #make it into a undirected networkx graph. This measures moduilarity on undirected version of graph!
partition = best_partition(modgraph) #find the partition with maximal modularity
modval = modularity(partition,modgraph) #calculate modularity with that partition
sleep(2)
if random == True:
random_modgraph = random_graph.to_undirected() #make it into a undirected networkx graph. This measures moduilarity on undirected version of graph!
random_partition = best_partition(random_modgraph) #find the partition with maximal modularity
random_modval = modularity(random_partition,random_modgraph) #calculate modularity with that partition
#option to use python nx to moralize and triangulate
if usenx == True:
moralized = moral_graph(graph)
if random == True:
random_moralized = moral_graph(random_graph)
else: #use Octave to run Matlab Bayes Net Toolbox, NOT SET UP FOR RANDOM GRAPHS YET.
moralized = nx.to_numpy_matrix(graph) # make nx graph into a simple matrix
ismoral = False
tries = 0
while ismoral == False and tries < 5: #sometimes oct2py takes too long to return I think
try:
moralized, moral_edges = octave.moralize(moralized)
ismoral = True
except:
ismoral = False
print 'Octave Error, trying again!'
tries = tries + 1
order = range(len(moralized)+1) # BNT needs order of nodes to triangulate
order = order[1:]
random_order = range(len(random_moralized)+1) # BNT needs order of nodes to triangulate
random_order = random_order[1:] # I think you have to shift the space because you are going from 0 index to 1 idndex
istriangulated = False
tries = 0
while istriangulated == False and tries < 5: #sometimes oct2py takes too long to return I think
try:
moralized = nx.to_numpy_matrix(moralized) # have to make it into matrix.
triangulated, cliques, fill_ins = octave.triangulate(moralized,order)
if random == True:
random_moralized = nx.to_numpy_matrix(random_moralized) # have to make it into matrix.
random_triangulated, random_cliques, random_fill_ins = octave.triangulate(random_moralized, random_order)
istriangulated = True
tries = 5
except:
istriangulated = False
print 'Octave Error, trying again!'
tries = tries + 1
#loop through cliques and get the size of them
cliquesizes = [] #empty array to keep clique sizes in
for x in cliques:
size = len(x)
cliquesizes.append(size)
maxclique = max(cliquesizes) #get the biggest clique
avgclique = np.mean(cliquesizes) # get the mean of the clique sizes
#do the same for random graph cliques
if random == True:
random_cliquesizes = [] #empty array to keep clique sizes in
#loop through cliques and get the size of them
for x in random_cliques:
size = len(x)
random_cliquesizes.append(size)
random_maxclique = max(random_cliquesizes) #get the biggest clique
random_avgclique = np.mean(random_cliquesizes) # get the mean of the clique sizes
end_time = time() #get end time
run_time = end_time - start_time #time how long all the took
#store the data!
if random == True:
data.append([nodegrowth, edgegrowth, modval, random_modval, maxclique, random_maxclique, avgclique, random_avgclique, run_time]) #store results
if random == False:
data.append([nodegrowth, edgegrowth, modval, maxclique,avgclique,run_time]) #store results
sparsemeasurements.remove(nodegrowth) #so we don't calculate clique size more than once!
if verbose:
print 'took ' + str(run_time) + ' to run last computation'
#this will always print, basic status update everytime clique size is measured.
if random == True:
print str(nodegrowth) + ' nodes, ', str(edgegrowth) + ' edges; ' + 'Modularity: ' + str(modval) + ', Random Modularity: ' +str(random_modval) + ', Largest Clique: ' + str(maxclique) + ', Largest Random Clique: ' + str(random_maxclique)
if random == False:
print str(nodegrowth) + ' nodes, ', str(edgegrowth) + ' edges; ' + 'Modularity: ' + str(modval) + ', Largest Clique: ' + str(maxclique)
#this will redraw the plot everytime a computation is done.
if plot == True:
if random == False:
df = pd.DataFrame(data, columns= ('nodegrowth','edgegrowth', 'modularity','maxclique','avgclique','run_time'))
plt.close()
fig = plt.figure(figsize=(24,16))
ax = fig.add_subplot(1,1,1)
ax2 = ax.twinx()
y1 = df[ploty] #user input, default to clique size
y2 = df[ploty2] #user input, default to modularity
x = df[plotx] #user input, default to nodes in graph.
ax.set_xlabel('%s in Graph' %(plotx),fontsize=20)
line1, = ax.plot(x, y1, label = ploty)
line2, = ax2.plot(x, y2, label = ploty2, color = 'red')
ax.set_ylabel(ploty,fontsize=20)
ax2.set_ylabel(ploty2, fontsize=20)
plt.suptitle('Graph Growth', fontsize=30)
plt.legend((line1,line2),(ploty , ploty2), loc='upper center', frameon=False, fontsize=20)
plt.show()
if random == True:
df = pd.DataFrame(data, columns= ('nodegrowth', 'edgegrowth', 'modval','random_modval', 'maxclique', 'random_maxclique', 'avgclique', 'random_avgclique', 'run_time'))
plt.close()
fig = plt.figure(figsize=(18,10))
ax = fig.add_subplot(1,1,1)
ax2 = ax.twinx()
ax3 = ax.twinx()
y1 = df[ploty]
y2 = df[ploty2]
y3def = str('random_'+ploty) # i just add random to whatever the user inputs as the y stuff they want to plot
y3 = df[y3def]
y4def = str('random_'+ploty2)
y4 = df[y4def]
x = df[plotx]
ax.set_xlabel('%s in Graph' %(plotx),fontsize=20)
line1, = ax.plot(x, y1, label = ploty, color = 'blue')
line2, = ax2.plot(x, y2, label = ploty2, color = 'green')
line3, = ax.plot(x,y3, label = y3def, color = 'red')
line4, = ax3.plot(x,y4, label = y4def, color = 'cyan')
ax.set_ylabel(ploty,fontsize=20)
ax2.set_ylabel(ploty2, fontsize=20)
plt.suptitle('Graph Growth', fontsize=30)
plt.legend((line1,line2,line3,line4), (ploty,ploty2,y3def,y4def),loc='upper center', frameon=False, fontsize=20)#
plt.show()
#draw the graph
if draw == True:
if drawgraph == 'triangulated':
G = nx.from_numpy_matrix(triangulated)
elif drawgraph == 'moralized':
G = nx.from_numpy_matrix(moralized)
elif drawgraph == 'directed':
G = graph
plt.close()
if drawspectral == True:
nx.draw_random(G, prog='neato')
else:
pos = nx.random_layout(G)
# find node near center (0.5,0.5)
dmin=1
ncenter=0
for n in pos:
x,y=pos[n]
d=(x-0.5)**2+(y-0.5)**2
if d<dmin:
ncenter=n
dmin=d
# color by path length from node near center
p=nx.single_source_shortest_path_length(G,ncenter)
plt.figure(figsize=(15,15))
plt.suptitle(drawgraph + ' graph')
nx.draw_networkx_edges(G,pos,nodelist=[ncenter],alpha=0.2)
nx.draw_networkx_nodes(G,pos,nodelist=p.keys(),
node_size=20,
node_color=p.values(),
cmap=plt.cm.Reds_r)
plt.xlim(-0.05,1.05)
plt.ylim(-0.05,1.05)
plt.axis('off')
plt.show()
if random == True:
df = pd.DataFrame(data, columns= ('nodegrowth', 'edgegrowth', 'modval','random_modval', 'maxclique', 'random_maxclique', 'avgclique', 'random_avgclique', 'run_time'))
if random == False:
df = pd.DataFrame(data, columns= ('nodegrowth','edgegrowth', 'modval','maxclique','avgclique','run_time'))
if getgraph == True:
return (graph, df)
else:
return(df)
data = []
if graph_db.size < 2:
print 'Your graph is empty. Please load a graph into the database.'
1/0
if sparse == True:
sparsemeasurements = [6,7,8,9,growthfactor]
measurements = np.logspace(1, (int(len(str(growthfactor))))-1, num_measurements)
for x in measurements:
sparsemeasurements.append(int(x))
else:
sparsemeasurements = range(2,growthfactor)
if directed:
graph = nx.DiGraph()
else:
graph = nx.graph()
nodegrowth = graph.size()
edgegrowth = 0
if wholegrowth == True:
growthfactor = len(nodes)
initial_node = random.choice(nodes.keys()) #randomly get a node to add from dictionary
if verbose:
print 'Starting from ' + initial_node
while nodegrowth < growthfactor:
if nodegrowth > 1:
possiblenodes = graph.nodes() # get all nodes in graph.
fromnode = random.choice(possiblenodes) #pick random node in graph to grow from.
edges = []
for node in data:
for edge in node['timelineItems']['edges']:
crnode = edge['node']
if not crnode['isCrossRepository']:
edges.append((node['number'], crnode['source']['number']))
edges
refcounts
refcounts.sum()
len(edges)
nodes
G = nx.graph()
import networkx as nx
G = nx.graph()
G = nx.Graph()
G
G.add_nodes(nodes)
G.add_nodes_from(nodes)
G.nodes
G.add_edges_from(edges)
G.nodes
nx.draw(G)
plt.show()
import matplotlib.pyplot as plt
plt.show()
nx.draw(G, node_size=100)
plt.show()
G
G.edges
def modMST(graph):
result = nx.graph()
sort = sorted(g.edges(data=True), key = lambda t: t[2].get('weight'),reverse=True)
def grow_graph(reverserandom=False,
outgoingrandom=False,
incomingrandom=False,
totalrandom=False,
usenx=True,
force_connected=True,
sparse=True,
plot=False,
directed=True,
randomgrowth=False,
wholegrowth=False,
growthfactor=100,
num_measurements=10,
verbose=True,
connected=True,
plotx='nodegrowth',
ploty='maxclique',
ploty2='modval',
drawgraph='triangulated',
draw=True,
drawspectral=True,
getgraph=True):
random = False #set as false, gets made to truth later if the person passes a type of random graph they want
#make sure user does not want to draw and plot at the same time.
if plot == True:
draw = False
if draw == True:
plot = False
#you cannot plot and draw at the same time
# initialize database
graph_db = database()
#get the pickled dictionary of nodes and node names (for the database) that is made while the csv file is loaded.
try:
nodes = open('nodes.p', 'r')
nodes = pickle.load(nodes)
except:
print 'Your graph is empty. Please load a graph into the database.'
1 / 0
data = []
if graph_db.size < 2:
print 'Your graph is empty. Please load a graph into the database.'
1 / 0
# this will grow the whole graph. Takes a while for big ones!
if wholegrowth == True:
growthfactor = len(nodes)
#make a list of all the nodes in the database to randomly choose from, but only if force connected graph option is false.
if force_connected == False:
possiblenodes = nodes
print 'Graph will not be fully connected, use "force_connected = True" if you want it to be fully connected'
# here we figure out at what points to measure if the user wants a spare measumement or to measure every time a node is added. This speeds up big graph growths a lot!
if sparse == True:
sparsemeasurements = [10, 15, growthfactor]
measurements = np.logspace(1, (int(len(str(growthfactor)))) - 1,
num_measurements)
for x in measurements:
sparsemeasurements.append(int(x))
else:
sparsemeasurements = range(2, growthfactor)
# this will actually only work for directed graph because we are moralizing, but I want to leave the option for later. Perhaps I can just skip moralization for undirected graphs.
if directed:
graph = nx.DiGraph()
else:
graph = nx.graph()
# pick an initial node to start from
initial_node = np.random.choice(
nodes.keys()) #randomly get a node to add from dictionary
if verbose:
print 'Starting from ' + initial_node
initial_used = 0 #keep track of how many times the initial node was used
nodegrowth = graph.number_of_nodes()
edgegrowth = graph.number_of_edges()
while nodegrowth < growthfactor: # make sure we aren't above how many nodes we want to measure in the graph
#start off finding a node to add to the graph.
if force_connected == True:
try:
possiblenodes = graph.nodes() # get all nodes in graph.
fromnode = Random.choice(
possiblenodes
) #pick random node in graph to grow from(add one of its nieghbors). This uses the random module, not np.random
if verbose:
print 'Using ' + str(fromnode) + ' to find new node'
except: #this is because you can't do random from one node at the start.
fromnode = initial_node
graph.add_node(fromnode)
if verbose:
print 'using initial node'
initial_used = initial_used + 1
if initial_used > 5:
1 / 0
fromnode = nodes[fromnode] #get DB version of the node.
#get all relationships in graph DB for that node so we can pick
new_node_rels = list(
graph_db.match(end_node=fromnode, bidirectional=True))
new_rel = Random.choice(
new_node_rels
) #randomly pick one of them, thus picking a node to add to graph. This uses the random module, not np.random
# is the new node a parent or child of the node we are growing from?
if new_rel.end_node == fromnode:
new_node = new_rel.start_node
if new_rel.start_node == fromnode:
new_node = new_rel.end_node
assert new_node != fromnode
if force_connected == False: # if not connected, we can just pick from the pickled dictionary of nodes in the database
new_node = np.random.choice(possiblenodes.values())
print 'adding' + str(new_node)
#add the nodes to the graph, connecting it to nodes in the graph that it is connected to.
# go through the list of edges that have the new node as a part of it, and only add the edge if they are between the new node and a node in the graph already.
rels = list(graph_db.match(
start_node=new_node)) #query graph for edges from that node
for edge in rels:
#get the string name of the node
newnodename = edge.start_node.get_properties()
newnodename = newnodename.get('name')
newnodename = newnodename.encode()
endnodename = edge.end_node.get_properties()
endnodename = endnodename.get('name')
endnodename = endnodename.encode()
if newnodename not in graph: #check to see if new node is in graph
graph.add_node(newnodename) # add if not
if verbose:
print 'added ' + str(newnodename)
if endnodename in graph: #check to see if end node is in graph
graph.add_edge(newnodename, endnodename) #add it if it is
if verbose:
print 'connected ' + newnodename + ' to ' + endnodename
rels = list(graph_db.match(
end_node=new_node)) #query graph for edges to that node
for edge in rels:
newnodename = edge.end_node.get_properties()
newnodename = newnodename.get('name')
newnodename = newnodename.encode()
startnodename = edge.start_node.get_properties()
startnodename = startnodename.get('name')
startnodename = startnodename.encode()
if newnodename not in graph: #check to see if new node is in graph
graph.add_node(newnodename) # add if not
if verbose:
print 'added ' + str(newnodename)
if startnodename in graph: #check to see if end node is in graph
graph.add_edge(startnodename, newnodename) #add it if it is
if verbose:
print 'connected ' + startnodename + ' to ' + newnodename
#measure the nodegrowth and edge growth
nodegrowth = len(graph.nodes())
edgegrowth = len(graph.edges())
if verbose:
print 'Graph has ' + str(nodegrowth) + ' nodes.'
if verbose:
print 'Graph has ' + str(edgegrowth) + ' edges.'
#here I make a few random graphs to compare the growth of the real graph. They are all the same size and same edges, but different controls.
# Reverse the direction of the edges. The in-degree distribution in this graph is power-law, but the out-degree is exponential tailed. So this is just a check that degree distribution is irrelevant.
if reverserandom == True:
random_graph = graph.reverse()
random = True
# Keep the number of outgoing links for each node the same, but randomly allocating their destinations. This should break modularity, put preserves out-degree.
if outgoingrandom == True:
random_graph = graph.copy()
for edge in random_graph.edges():
parent = edge[0]
child = edge[1]
random_graph.remove_edge(parent, child)
newchild = parent
while newchild == parent: #so that we don't get a self loop.
newchild = np.random.choice(graph.nodes())
random_graph.add_edge(parent, newchild)
random = True
# Same thing, but this time fixing the number of incoming links and randomly allocating their origins. Likewise, but preserves in-degree.
if incomingrandom == True:
random_graph = graph.copy()
for edge in random_graph.edges():
parent = edge[0]
child = edge[1]
random_graph.remove_edge(parent, child)
newparent = child
while newparent == child:
newparent = np.random.choice(graph.nodes())
random_graph.add_edge(newparent, child)
random = True
#gives a graph picked randomly out of the set of all graphs with n nodes and m edges. This preserves overall degree, and number of nodes/edges, but is completeley random to outdegree/indegree.
if totalrandom == True:
numrandomedges = graph.number_of_edges()
numrandomnodes = graph.number_of_nodes()
random_graph = nx.dense_gnm_random_graph(numrandomnodes,
numrandomedges)
random_graph = random_graph.to_directed()
random = True
#here is where we measure everything about the graph
if nodegrowth in sparsemeasurements: #we only measure every now and then in sparse.
start_time = time()
if nodegrowth > 5:
modgraph = nx.Graph(
graph
) #make it into a undirected networkx graph. This measures moduilarity on undirected version of graph!
partition = best_partition(
modgraph) #find the partition with maximal modularity
modval = modularity(
partition,
modgraph) #calculate modularity with that partition
sleep(2)
if random == True:
random_modgraph = random_graph.to_undirected(
) #make it into a undirected networkx graph. This measures moduilarity on undirected version of graph!
random_partition = best_partition(
random_modgraph
) #find the partition with maximal modularity
random_modval = modularity(
random_partition, random_modgraph
) #calculate modularity with that partition
#option to use python nx to moralize and triangulate
if usenx == True:
moralized = moral_graph(graph)
if random == True:
random_moralized = moral_graph(random_graph)
else: #use Octave to run Matlab Bayes Net Toolbox, NOT SET UP FOR RANDOM GRAPHS YET.
moralized = nx.to_numpy_matrix(
graph) # make nx graph into a simple matrix
ismoral = False
tries = 0
while ismoral == False and tries < 5: #sometimes oct2py takes too long to return I think
try:
moralized, moral_edges = octave.moralize(moralized)
ismoral = True
except:
ismoral = False
print 'Octave Error, trying again!'
tries = tries + 1
order = range(len(moralized) +
1) # BNT needs order of nodes to triangulate
order = order[1:]
random_order = range(len(random_moralized) +
1) # BNT needs order of nodes to triangulate
random_order = random_order[
1:] # I think you have to shift the space because you are going from 0 index to 1 idndex
istriangulated = False
tries = 0
while istriangulated == False and tries < 5: #sometimes oct2py takes too long to return I think
try:
moralized = nx.to_numpy_matrix(
moralized) # have to make it into matrix.
triangulated, cliques, fill_ins = octave.triangulate(
moralized, order)
if random == True:
random_moralized = nx.to_numpy_matrix(
random_moralized) # have to make it into matrix.
random_triangulated, random_cliques, random_fill_ins = octave.triangulate(
random_moralized, random_order)
istriangulated = True
tries = 5
except:
istriangulated = False
print 'Octave Error, trying again!'
tries = tries + 1
#loop through cliques and get the size of them
cliquesizes = [] #empty array to keep clique sizes in
for x in cliques:
size = len(x)
cliquesizes.append(size)
maxclique = max(cliquesizes) #get the biggest clique
avgclique = np.mean(
cliquesizes) # get the mean of the clique sizes
#do the same for random graph cliques
if random == True:
random_cliquesizes = [] #empty array to keep clique sizes in
#loop through cliques and get the size of them
for x in random_cliques:
size = len(x)
random_cliquesizes.append(size)
random_maxclique = max(
random_cliquesizes) #get the biggest clique
random_avgclique = np.mean(
random_cliquesizes) # get the mean of the clique sizes
end_time = time() #get end time
run_time = end_time - start_time #time how long all the took
#store the data!
if random == True:
data.append([
nodegrowth, edgegrowth, modval, random_modval, maxclique,
random_maxclique, avgclique, random_avgclique, run_time
]) #store results
if random == False:
data.append([
nodegrowth, edgegrowth, modval, maxclique, avgclique,
run_time
]) #store results
sparsemeasurements.remove(
nodegrowth) #so we don't calculate clique size more than once!
if verbose:
print 'took ' + str(run_time) + ' to run last computation'
#this will always print, basic status update everytime clique size is measured.
if random == True:
print str(nodegrowth) + ' nodes, ', str(
edgegrowth) + ' edges; ' + 'Modularity: ' + str(
modval) + ', Random Modularity: ' + str(
random_modval) + ', Largest Clique: ' + str(
maxclique) + ', Largest Random Clique: ' + str(
random_maxclique)
if random == False:
print str(nodegrowth) + ' nodes, ', str(
edgegrowth) + ' edges; ' + 'Modularity: ' + str(
modval) + ', Largest Clique: ' + str(maxclique)
#this will redraw the plot everytime a computation is done.
if plot == True:
if random == False:
df = pd.DataFrame(data,
columns=('nodegrowth', 'edgegrowth',
'modularity', 'maxclique',
'avgclique', 'run_time'))
plt.close()
fig = plt.figure(figsize=(24, 16))
ax = fig.add_subplot(1, 1, 1)
ax2 = ax.twinx()
y1 = df[ploty] #user input, default to clique size
y2 = df[ploty2] #user input, default to modularity
x = df[plotx] #user input, default to nodes in graph.
ax.set_xlabel('%s in Graph' % (plotx), fontsize=20)
line1, = ax.plot(x, y1, label=ploty)
line2, = ax2.plot(x, y2, label=ploty2, color='red')
ax.set_ylabel(ploty, fontsize=20)
ax2.set_ylabel(ploty2, fontsize=20)
plt.suptitle('Graph Growth', fontsize=30)
plt.legend((line1, line2), (ploty, ploty2),
loc='upper center',
frameon=False,
fontsize=20)
plt.show()
if random == True:
df = pd.DataFrame(data,
columns=('nodegrowth', 'edgegrowth',
'modval', 'random_modval',
'maxclique', 'random_maxclique',
'avgclique', 'random_avgclique',
'run_time'))
plt.close()
fig = plt.figure(figsize=(18, 10))
ax = fig.add_subplot(1, 1, 1)
ax2 = ax.twinx()
ax3 = ax.twinx()
y1 = df[ploty]
y2 = df[ploty2]
y3def = str(
'random_' + ploty
) # i just add random to whatever the user inputs as the y stuff they want to plot
y3 = df[y3def]
y4def = str('random_' + ploty2)
y4 = df[y4def]
x = df[plotx]
ax.set_xlabel('%s in Graph' % (plotx), fontsize=20)
line1, = ax.plot(x, y1, label=ploty, color='blue')
line2, = ax2.plot(x, y2, label=ploty2, color='green')
line3, = ax.plot(x, y3, label=y3def, color='red')
line4, = ax3.plot(x, y4, label=y4def, color='cyan')
ax.set_ylabel(ploty, fontsize=20)
ax2.set_ylabel(ploty2, fontsize=20)
plt.suptitle('Graph Growth', fontsize=30)
plt.legend((line1, line2, line3, line4),
(ploty, ploty2, y3def, y4def),
loc='upper center',
frameon=False,
fontsize=20) #
plt.show()
#draw the graph
if draw == True:
if drawgraph == 'triangulated':
G = nx.from_numpy_matrix(triangulated)
elif drawgraph == 'moralized':
G = nx.from_numpy_matrix(moralized)
elif drawgraph == 'directed':
G = graph
plt.close()
if drawspectral == True:
nx.draw_random(G, prog='neato')
else:
pos = nx.random_layout(G)
# find node near center (0.5,0.5)
dmin = 1
ncenter = 0
for n in pos:
x, y = pos[n]
d = (x - 0.5)**2 + (y - 0.5)**2
if d < dmin:
ncenter = n
dmin = d
# color by path length from node near center
p = nx.single_source_shortest_path_length(G, ncenter)
plt.figure(figsize=(15, 15))
plt.suptitle(drawgraph + ' graph')
nx.draw_networkx_edges(G,
pos,
nodelist=[ncenter],
alpha=0.2)
nx.draw_networkx_nodes(G,
pos,
nodelist=p.keys(),
node_size=20,
node_color=p.values(),
cmap=plt.cm.Reds_r)
plt.xlim(-0.05, 1.05)
plt.ylim(-0.05, 1.05)
plt.axis('off')
plt.show()
if random == True:
df = pd.DataFrame(data,
columns=('nodegrowth', 'edgegrowth', 'modval',
'random_modval', 'maxclique',
'random_maxclique', 'avgclique',
'random_avgclique', 'run_time'))
if random == False:
df = pd.DataFrame(data,
columns=('nodegrowth', 'edgegrowth', 'modval',
'maxclique', 'avgclique', 'run_time'))
if getgraph == True:
return (graph, df)
else:
return (df)