def make_final_net():
A = BayesNode(0, 2, name='A')
B = BayesNode(1, 2, name='B')
C = BayesNode(2, 2, name='C')
D = BayesNode(3, 2, name='D')
E = BayesNode(4, 2, name='E')
A.add_child(C)
C.add_parent(A)
B.add_child(C)
C.add_parent(B)
B.add_child(D)
D.add_parent(B)
C.add_child(E)
E.add_parent(C)
D.add_child(E)
E.add_parent(D)
nodes = [A, B, C, D, E]
net = BayesNet(nodes)
A = net.get_node_by_name('A')
B = net.get_node_by_name('B')
C = net.get_node_by_name('C')
D = net.get_node_by_name('D')
E = net.get_node_by_name('E')
A_distribution = DiscreteDistribution(A)
index = A_distribution.generate_index([], [])
A_distribution[index] = [0.4, 0.6]
A.set_dist(A_distribution)
B_distribution = DiscreteDistribution(B)
index = B_distribution.generate_index([], [])
B_distribution[index] = [0.8, 0.2]
B.set_dist(B_distribution)
print B, D
dist = zeros([B.size(), D.size()], dtype=float32)
dist[0, :] = [0.87, 0.13]
dist[1, :] = [0.24, 0.76]
D_distribution = ConditionalDiscreteDistribution(nodes=[B, D], table=dist)
D.set_dist(D_distribution)
dist = zeros([A.size(), B.size(), C.size()], dtype=float32)
dist[0, 0, :] = [0.85, 0.15]
dist[0, 1, :] = [0.68, 0.32]
dist[1, 0, :] = [0.16, 0.84]
dist[1, 1, :] = [0.05, 0.95]
C_distribution = ConditionalDiscreteDistribution(nodes=[A, B, C],
table=dist)
C.set_dist(C_distribution)
dist = zeros([C.size(), D.size(), E.size()], dtype=float32)
dist[0, 0, :] = [0.8, 0.2]
dist[0, 1, :] = [0.37, 0.63]
dist[1, 0, :] = [0.08, 0.92]
dist[1, 1, :] = [0.07, 0.93]
E_distribution = ConditionalDiscreteDistribution(nodes=[C, D, E],
table=dist)
E.set_dist(E_distribution)
engine = EnumerationEngine(net)
# engine1.evidence[wear] = False
# engine1.evidence[weap] = True
engine.evidence[A] = True
engine.evidence[C] = True
Q = engine.marginal(E)[0]
index = Q.generate_index([True], range(Q.nDims))
prob = Q[index]
print prob
def make_exam_net():
H = BayesNode(0, 2, name='H')
G = BayesNode(1, 2, name='G')
B = BayesNode(2, 2, name='B')
O = BayesNode(3, 2, name='O')
D = BayesNode(4, 2, name='D')
C = BayesNode(5, 2, name='C')
H.add_child(B)
B.add_parent(H)
B.add_parent(G)
G.add_child(B)
G.add_child(O)
O.add_parent(G)
O.add_child(D)
O.add_child(C)
D.add_parent(O)
C.add_parent(O)
B.add_child(D)
D.add_parent(B)
nodes = [B, C, D, G, H, O]
net = BayesNet(nodes)
sci = BayesNode(0, 2, name='sci')
inf = BayesNode(1, 2, name='inf')
weap = BayesNode(2, 2, name='weap')
car = BayesNode(3, 2, name='car')
wear = BayesNode(4, 2, name='wear')
sci.add_child(weap)
weap.add_parent(sci)
inf.add_child(weap)
weap.add_parent(inf)
weap.add_child(car)
car.add_parent(weap)
weap.add_child(wear)
wear.add_parent(weap)
nodes1 = [sci, weap, wear, inf, car]
net1 = BayesNet(nodes1)
B = net.get_node_by_name('B')
C = net.get_node_by_name('C')
D = net.get_node_by_name('D')
G = net.get_node_by_name('G')
H = net.get_node_by_name('H')
O = net.get_node_by_name('O')
H_distribution = DiscreteDistribution(H)
index = H_distribution.generate_index([], [])
H_distribution[index] = [0.6, 0.4]
H.set_dist(H_distribution)
G_distribution = DiscreteDistribution(G)
index = G_distribution.generate_index([], [])
G_distribution[index] = [0.75, 0.25]
G.set_dist(G_distribution)
dist = zeros([O.size(), C.size()], dtype=float32)
dist[0, :] = [0.55, 0.45]
dist[1, :] = [0.75, 0.25]
C_distribution = ConditionalDiscreteDistribution(nodes=[O, C], table=dist)
C.set_dist(C_distribution)
dist = zeros([G.size(), O.size()], dtype=float32)
dist[0, :] = [0.55, 0.45]
dist[1, :] = [0.45, 0.55]
O_distribution = ConditionalDiscreteDistribution(nodes=[G, O], table=dist)
O.set_dist(O_distribution)
dist = zeros([B.size(), O.size(), D.size()], dtype=float32)
dist[0, 0, :] = [0.72, 0.28]
dist[0, 1, :] = [0.38, 0.62]
dist[1, 0, :] = [0.85, 0.15]
dist[1, 1, :] = [0.65, 0.35]
D_distribution = ConditionalDiscreteDistribution(nodes=[B, O, D],
table=dist)
D.set_dist(D_distribution)
dist = zeros([H.size(), G.size(), B.size()], dtype=float32)
dist[0, 0, :] = [0.92, 0.08]
dist[0, 1, :] = [0.75, 0.25]
dist[1, 0, :] = [0.55, 0.45]
dist[1, 1, :] = [0.35, 0.65]
B_distribution = ConditionalDiscreteDistribution(nodes=[H, G, B],
table=dist)
B.set_dist(B_distribution)
sci = net1.get_node_by_name('sci')
weap = net1.get_node_by_name('weap')
wear = net1.get_node_by_name('wear')
inf = net1.get_node_by_name('inf')
car = net1.get_node_by_name('car')
sci_distribution = DiscreteDistribution(sci)
index = sci_distribution.generate_index([], [])
sci_distribution[index] = [0.2, 0.8]
sci.set_dist(sci_distribution)
inf_distribution = DiscreteDistribution(inf)
index = inf_distribution.generate_index([], [])
inf_distribution[index] = [0.4, 0.6]
inf.set_dist(inf_distribution)
dist = zeros([sci.size(), inf.size(), weap.size()], dtype=float32)
dist[0, 0, :] = [0.6, 0.4]
dist[0, 1, :] = [0.8, 0.2]
dist[1, 0, :] = [0.1, 0.9]
dist[1, 1, :] = [0.3, 0.7]
weap_distribution = ConditionalDiscreteDistribution(nodes=[sci, inf, weap],
table=dist)
weap.set_dist(weap_distribution)
dist = zeros([weap.size(), wear.size()], dtype=float32)
dist[0, :] = [0.88, 0.12]
dist[1, :] = [0.15, 0.85]
wear_distribution = ConditionalDiscreteDistribution(nodes=[weap, wear],
table=dist)
wear.set_dist(wear_distribution)
dist = zeros([weap.size(), car.size()], dtype=float32)
dist[0, :] = [0.75, 0.25]
dist[1, :] = [0.55, 0.45]
car_distribution = ConditionalDiscreteDistribution(nodes=[weap, car],
table=dist)
car.set_dist(car_distribution)
## engine = JunctionTreeEngine(net)
# engine = EnumerationEngine(net)
## engine.evidence[B] = True
# Q = engine.marginal(C)[0]
# index = Q.generate_index([True], range(Q.nDims))
# prob = Q[index]
# print "Thr ptob of O = T given B = T is ", prob
engine1 = EnumerationEngine(net1)
engine1.evidence[wear] = False
engine1.evidence[weap] = True
# engine1.evidence[sci] = False
Q = engine1.marginal(car)[0]
index = Q.generate_index([True], range(Q.nDims))
prob = Q[index]
print prob