def _init_network(self, name, nodes_def):
"""
This responsible for the initilisation of the network
"""
nodes = {}
connections = []
network = BNet(name)
for (nodename, isdiscrete, numstates, leafnode) in nodes_def:
nodes[nodename] = network.add_v(BVertex(nodename, isdiscrete,
numstates))
# TODO: Find a way to improve this and avoid to have to loop a
# second time
# reason is because BNnodes[leafnode] is not sure to be there when
# going through the first loop
for (nodename, isdiscrete, numstates, leafnode) in nodes_def:
if type(leafnode) == type([]):
for r in leafnode:
if r != None:
connections.append((nodes[nodename], nodes[r]))
elif leafnode != None:
connections.append((nodes[nodename], nodes[leafnode]))
else:
#do nothing
pass
for ep in connections:
network.add_e(ep)
self._network = network
# Let's not forget to initialize the distribution
self._network.finalize()
def main():
"""
The main function
"""
# create the network
graph = BNet( 'Water Sprinkler Bayesian Network' )
b,c = [graph.add_v( BVertex( nm, False, 1 ) ) for nm in 'b c'.split()]
a = graph.add_v(BVertex('a',True,2))
for start, end in [( a, b ), ( b, c )]:
graph.add_e( DirEdge( len( graph.e ), start, end ) )
print graph
# finalize the bayesian network once all edges have been added
graph.init_distributions()
# fill in the parameters
a.distribution.set_parameters([0.7,0.3])
b.distribution.set_parameters(mu=[2.0,0.0], sigma=[1.0,1.0])
c.distribution.set_parameters(mu=2.0, sigma=1.0, wi=1.0)
case = graph.sample(1) #---TODO: correct this bug...
print case
# NOTE : for the moment only MCMCEngine can work for continuous networks
engine = MCMCEngine(graph)
res = engine.marginalise_all()
for res in res.values():
print res
def main():
"""
Simply the main function. We watn to be clean
"""
# create the network
graph = BNet( 'Water Sprinkler Bayesian Network' )
node_a, node_b, node_c = [graph.add_v(BVertex(nm, False, 1))
for nm in 'a b c'.split()]
for start, end in [(node_a, node_b), (node_b, node_c)]:
graph.add_e(DirEdge(len(graph.e), start, end ))
print graph
# finalize the bayesian network once all edges have been added
graph.init_distributions()
# fill in the parameters
node_a.distribution.set_parameters(mu=1.0, sigma=0.5)
node_b.distribution.set_parameters(mu=2.0, sigma=1.0, wi=2.0)
node_c.distribution.set_parameters(mu=2.0, sigma=1.0, wi=1.0)
# NOTE : for the moment only MCMCEngine can work for continuous networks
engine = MCMCEngine(graph)
res = engine.marginalise_all()
for res in res.values():
print res,'\n'
def main():
"""
The main function
"""
# create the network
graph = BNet('A Simple Bayesian Network')
node_a, node_b = [graph.add_v(BVertex(nm, True, 2))
for nm in 'a b'.split()]
graph.add_e(DirEdge(len(graph.e), node_a, node_b))
graph_copy = deepcopy(graph)
print graph
# finalize the bayesian network once all edges have been added
graph.init_distributions()
print node_a.distribution.cpt, node_b.distribution.cpt
print '+' * 20, '\n'
graph_copy.init_distributions()
print graph_copy.all_v[0].distribution.cpt
engine = learning.MLLearningEngine(graph)
try:
cases = engine.read_file('test.xls')
except ImportError:
print "No support for excel. Install the xlrd module"
print "This test does not continue"
return
print 'cases:', cases
start_time = time()
engine.learn_ml_params(cases, 0)
print 'Learned from 4 cases in %1.3f secs' % ((time() - start_time))
# print the parameters
for vertex in graph.all_v:
print vertex.distribution,'\n'
def main():
"""
This is the main function. It simply create a baysian network and
return it.
"""
# We start by creating the network
graph = BNet( 'Water Sprinkler Bayesian Network' )
cloudy, sprinkler, rainy, wet = [graph.add_v(BVertex(nm, True, 2))
for nm in 'c s r w'.split()]
for edge_pair in [(cloudy, rainy), (cloudy, sprinkler),
(rainy, wet), (sprinkler, wet)]:
graph.add_e(DirEdge(len(graph.e), edge_pair[0], edge_pair[1]))
print graph
# finalize the bayesian network once all edges have been added
graph.init_distributions()
# c | Pr(c)
#---+------
# 0 | 0.5
# 1 | 0.5
cloudy.set_distribution_parameters([0.5, 0.5])
# c s | Pr(s|c)
#-----+--------
# 0 0 | 0.5
# 1 0 | 0.9
# 0 1 | 0.5
# 1 1 | 0.1
sprinkler.set_distribution_parameters([0.5, 0.9, 0.5, 0.1])
# c r | Pr(r|c)
#-----+--------
# 0 0 | 0.8
# 1 0 | 0.2
# 0 1 | 0.2
# 1 1 | 0.8
rainy.set_distribution_parameters([0.8, 0.2, 0.2, 0.8])
# s r w | Pr(w|c,s)
#-------+------------
# 0 0 0 | 1.0
# 1 0 0 | 0.1
# 0 1 0 | 0.1
# 1 1 0 | 0.01
# 0 0 1 | 0.0
# 1 0 1 | 0.9
# 0 1 1 | 0.9
# 1 1 1 | 0.99
# to verify the order of variables use :
#>>> w.distribution.names_list
#['w','s','r']
# we can also set up the variables of a cpt with the following way
# again the order is w.distribution.names_list
wet.distribution[:, 0, 0] = [0.99, 0.01]
wet.distribution[:, 0, 1] = [0.1, 0.9]
wet.distribution[:, 1, 0] = [0.1, 0.9]
wet.distribution[:, 1, 1] = [0.0, 1.0]
# or even this way , using a dict:
#w.distribution[{'s':0,'r':0}]=[0.99, 0.01]
#w.distribution[{'s':0,'r':1}]=[0.1, 0.9]
#w.distribution[{'s':1,'r':0}]=[0.1, 0.9]
#w.distribution[{'s':1,'r':1}]=[0.0, 1.0]
return graph
