def _init_BN(self, BNnodesdef):
BNnodes = {}
BNconnections = []
BN = BNet()
# Save the nodesdef for future use (saving)
self._BNDef = BNnodesdef
for node in BNnodesdef:
(nodename , isdiscrete, numstates, leafnode) = node
BNnodes[nodename] = BN.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 node in BNnodesdef:
(nodename , isdiscrete, numstates, leafnode) = node
if type(leafnode)==type([]):
for r in leafnode:
if r!=None: BNconnections.append( (BNnodes[nodename], BNnodes[r]) )
elif leafnode!=None:
BNconnections.append( (BNnodes[nodename], BNnodes[leafnode]) )
else:
#do nothing
pass
for ep in BNconnections:
BN.add_e( DirEdge( len( BN.e ), *ep ) )
# Ok our Bnet has been created, let's save it in the controller
self._BN = BN
# Let's not forget to initialize the distribution
self._BN.InitDistributions()
def _init_BN(self, BNName, BNnodesdef):
BNnodes = {}
BNconnections = []
BN = BNet(BNName)
# Save the nodesdef for future use (saving)
self._BNDef = BNnodesdef
for node in BNnodesdef:
(nodename, isdiscrete, numstates, leafnode) = node
BNnodes[nodename] = BN.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 node in BNnodesdef:
(nodename, isdiscrete, numstates, leafnode) = node
if type(leafnode) == type([]):
for r in leafnode:
if r != None:
BNconnections.append((BNnodes[nodename], BNnodes[r]))
elif leafnode != None:
BNconnections.append((BNnodes[nodename], BNnodes[leafnode]))
else:
#do nothing
pass
for ep in BNconnections:
BN.add_e(DirEdge(len(BN.e), *ep))
# Ok our Bnet has been created, let's save it in the controller
self._BN = BN
# Let's not forget to initialize the distribution
self._BN.InitDistributions()
def CreateOpenBayesNetwork(network):
G = BNet( network.name )
node_dict = {}
for node in network.nodes.all():
pynode = BVertex( node.name, True, node.states.count() )
node_dict[node.id] = pynode
G.add_v(pynode)
for edge in network.edges.all():
pyedge = DirEdge( len( G.e ),node_dict[edge.parent_node.id] , node_dict[edge.child_node.id])
G.add_e( pyedge )
G.InitDistributions()
for node in network.nodes.all():
if not node.is_root():
result = node.get_indexed_value_sets()[0]
index = node.get_indexed_value_sets()[1]
for i in range(len(index)):
index_dict = {}
check_dict = {}
result_i_parents = result[i][0]
values_i = result[i][1]
index_i_parents = index[i][0]
for j in range(len(index_i_parents)):
index_dict[result_i_parents[j].node.name] = index_i_parents[j]
check_dict[result_i_parents[j].node.name] = result_i_parents[j].name
node_dict[node.id].distribution[index_dict] = [v.value for v in values_i]
print "\n"
else:
node_dict[node.id].setDistributionParameters([state.probability for state in node.states.all()])
return (G,node_dict)
network = BNet('Asia Bayesian Network')
# Create a discrete node for all nodes with 2 states
visit_to_asia = network.add_v(BVertex('Visit to Asia', True, 2)) # TRUE used to indicate the data is discrete as oppose to continuous
smoking = network.add_v(BVertex('Smoking', True, 2))
tuberculosis = network.add_v(BVertex('Tuberculosis', True, 2))
lung_cancer = network.add_v(BVertex('Lung Cancer', True, 2))
bronchitis = network.add_v(BVertex('Bronchitis', True, 2))
tub_or_cancer = network.add_v(BVertex('Tuberculosis or Cancer', True, 2))
xray_result = network.add_v(BVertex('X-Ray Result', True, 2))
dyspnea = network.add_v(BVertex('Dyspnea', True, 2))
# Connect the nodes
# V -> T
network.add_e(DirEdge(len(network.e), visit_to_asia, tuberculosis))
# T -> TC
network.add_e(DirEdge(len(network.e), tuberculosis, tub_or_cancer))
# TC -> X
network.add_e(DirEdge(len(network.e), tub_or_cancer, xray_result))
# TC -> D
network.add_e(DirEdge(len(network.e), tub_or_cancer, dyspnea))
# S -> LC
network.add_e(DirEdge(len(network.e), smoking, lung_cancer))
# S -> B
network.add_e(DirEdge(len(network.e), smoking, bronchitis))
A
/
B
\
C
all edges point downwards
A,B and C are univariate gaussian distributions
"""
from OpenBayes import BNet, BVertex, DirEdge, MCMCEngine
# create the network
G = BNet( 'Water Sprinkler Bayesian Network' )
a,b,c = [G.add_v( BVertex( nm, False, 1 ) ) for nm in 'a b c'.split()]
for ep in [( a, b ), ( b, c )]:
G.add_e( DirEdge( len( G.e ), *ep ) )
print G
# finalize the bayesian network once all edges have been added
G.InitDistributions()
# fill in the parameters
a.distribution.setParameters(mu=1.0, sigma=0.5)
b.distribution.setParameters(mu=2.0, sigma=1.0, wi=2.0)
c.distribution.setParameters(mu=2.0, sigma=1.0, wi=1.0)
# NOTE : for the moment only MCMCEngine can work for continuous networks
ie = MCMCEngine(G)
res = ie.MarginaliseAll()
network = BNet('Asia Bayesian Network')
# Create a discrete node for all nodes with 2 states
visit_to_asia = network.add_v(BVertex('Visit to Asia', True, 2))
smoking = network.add_v(BVertex('Smoking', True, 2))
tuberculosis = network.add_v(BVertex('Tuberculosis', True, 2))
lung_cancer = network.add_v(BVertex('Lung Cancer', True, 2))
bronchitis = network.add_v(BVertex('Bronchitis', True, 2))
tub_or_cancer = network.add_v(BVertex('Tuberculosis or Cancer', True, 2))
xray_result = network.add_v(BVertex('X-Ray Result', True, 2))
dyspnea = network.add_v(BVertex('Dyspnea', True, 2))
# Connect the nodes
# V -> T
network.add_e(DirEdge(len(network.e), visit_to_asia, tuberculosis))
# T -> TC
network.add_e(DirEdge(len(network.e), tuberculosis, tub_or_cancer))
# TC -> X
network.add_e(DirEdge(len(network.e), tub_or_cancer, xray_result))
# TC -> D
network.add_e(DirEdge(len(network.e), tub_or_cancer, dyspnea))
# S -> LC
network.add_e(DirEdge(len(network.e), smoking, lung_cancer))
# S -> B
network.add_e(DirEdge(len(network.e), smoking, bronchitis))
# create discrete node for all nodes with 2 states??
road_Density = network.add_v(BVertex('Road Density', True, 3))
unregulated_Hunting_Rate = network.add_v(BVertex('Unregulated Hunting Rate', True, 3))
cougar_risk_level = network.add_v(BVertex('Cougar Risk Level', True, 3))
wolfPresence = network.add_v(BVertex('Wolf Presence', True, 2))
predation_risk = network.add_v(BVertex('Predation Risk', True, 3))
population_hazard = network.add_v(BVertex('Population Hazard', True, 5))
proportion_of_lu_in_UWR = network.add_v(BVertex('Proportion of LU in UWR', True, 4))
uwr_hazard_rating = network.add_v(BVertex('UWR Hazard Rating', True, 5))
habitat_hazard_rating = network.add_v(BVertex('Habitat Hazard Rating', True, 5))
muleDeerHazardRating = network.add_v(BVertex('Mule Deer Hazard Rating', True, 5))
regulated_Hunting_Rate = network.add_v(BVertex('Regulated Hunting Rate', True, 3))
# now describe the connections
# population side of model
network.add_e(DirEdge(len(network.e), road_Density, unregulated_Hunting_Rate))
network.add_e(DirEdge(len(network.e), road_Density, predation_risk))
network.add_e(DirEdge(len(network.e), cougar_risk_level, predation_risk))
network.add_e(DirEdge(len(network.e), wolfPresence, predation_risk))
network.add_e(DirEdge(len(network.e), predation_risk, population_hazard))
network.add_e(DirEdge(len(network.e), unregulated_Hunting_Rate, population_hazard))
network.add_e(DirEdge(len(network.e), regulated_Hunting_Rate, population_hazard))
network.add_e(DirEdge(len(network.e), population_hazard, muleDeerHazardRating))
# habitat side of model
network.add_e(DirEdge(len(network.e), proportion_of_lu_in_UWR, habitat_hazard_rating))
network.add_e(DirEdge(len(network.e), uwr_hazard_rating, habitat_hazard_rating))
network.add_e(DirEdge(len(network.e), habitat_hazard_rating, muleDeerHazardRating))
network.InitDistributions()
# <1km / 1-2km / 2-3km
logging.getLogger("").setLevel(logging.CRITICAL)
import numpy
network = BNet('Number Guessing')
# Create a discrete node for all nodes with 2 states
numbers = network.add_v(BVertex('numbers', True, 10))
bigger_3 = network.add_v(BVertex('Is your number bigger than 3?', True, 2))
bigger_5 = network.add_v(BVertex('bigger_5', True, 2))
IsTheNumberBiggerThan7 = network.add_v(BVertex('IsTheNumberBiggerThan7', True, 2))
prime = network.add_v(BVertex('prime', True, 2))
even = network.add_v(BVertex('even', True, 2))
odd = network.add_v(BVertex('odd', True, 2))
network.add_e(DirEdge(len(network.e), numbers, bigger_3))
network.add_e(DirEdge(len(network.e), numbers, bigger_5))
network.add_e(DirEdge(len(network.e), numbers, IsTheNumberBiggerThan7))
network.add_e(DirEdge(len(network.e), numbers, even))
network.add_e(DirEdge(len(network.e), numbers, prime))
network.add_e(DirEdge(len(network.e), even, odd))
# Show the network
print network
# Initialize the distributions
network.InitDistributions()
# Set distributions for start nodes
numbers.setDistributionParameters([0.1]*10)
import numpy
network = BNet('Number Guessing')
# Create a discrete node for all nodes with 2 states
numbers = network.add_v(BVertex('numbers', True, 10))
bigger_3 = network.add_v(BVertex('Is your number bigger than 3?', True, 2))
bigger_5 = network.add_v(BVertex('bigger_5', True, 2))
IsTheNumberBiggerThan7 = network.add_v(
BVertex('IsTheNumberBiggerThan7', True, 2))
prime = network.add_v(BVertex('prime', True, 2))
even = network.add_v(BVertex('even', True, 2))
odd = network.add_v(BVertex('odd', True, 2))
network.add_e(DirEdge(len(network.e), numbers, bigger_3))
network.add_e(DirEdge(len(network.e), numbers, bigger_5))
network.add_e(DirEdge(len(network.e), numbers, IsTheNumberBiggerThan7))
network.add_e(DirEdge(len(network.e), numbers, even))
network.add_e(DirEdge(len(network.e), numbers, prime))
network.add_e(DirEdge(len(network.e), even, odd))
# Show the network
print network
# Initialize the distributions
network.InitDistributions()
# Set distributions for start nodes
numbers.setDistributionParameters([0.1] * 10)
