class Network(object):
def __init__(self, name, lic=None, file=None):
if file is None:
self.name = name
self.ntc = Netica()
self.env = self.ntc.newenv(lic)
self.ntc.initenv(self.env)
self.net = self.ntc.newnet(name, self.env)
self.nodes = []
else:
self.name = name
self.ntc = Netica()
self.env = self.ntc.newenv(lic)
self.ntc.initenv(self.env)
self.net = self.ntc.opennet(self.env, file)
nodelist_p = self.ntc.getnetnodes(self.net)
numnode = self.ntc.lengthnodelist(nodelist_p)
self.nodes = []
for i in range(numnode):
nodei_p = self.ntc.nthnode(nodelist_p, i)
nodename = self.ntc.getnodename(nodei_p)
statename = 'init'; statenames = []; istate = 0
while statename!='error':
statename = self.ntc.getnodestatename(nodei_p, istate)
statenames.append(statename)
istate += 1
statenames = statenames[:-1]
# default: no parents and continuous: therefore not the original nodes
nodei = Node(nodename, parents=[], rvname='continuous')
nodei.set_node_ptr(nodei_p)
nodei.set_node_state_name(statenames)
self.nodes.append(nodei)
def add_nodes(self, nodes):
for node in nodes:
if (node.nodekind==NATURE_NODE) and (node.cpt is None or (node.cpt.size == 0)):
print "assign {} cpt first before add to network".format(node.name)
#nodeptr = self.ntc.newnode(node.name, node.cpt.shape[0], self.net)
#node.set_pointer(nodeptr)
#node.set_net(self)
node.add_to_net(self)
self.nodes.append(node)
def add_link(self):
for node in self.nodes:
if node.parents is not None:
for parentNode in node.parents:
self.ntc.addlink(parent=parentNode.ptr, child=node.ptr)
def define_nodes(self):
"""define nodes can only be implemented after all links are added
"""
for node in self.nodes:
node.define()
def compile_net(self):
self.ntc.compilenet(self.net)
def set_autoupdate(self):
self.ntc.setautoupdate(self.net)
def get_node_beliefs(self, node):
beliefs32 = self.ntc.getnodebeliefs(node.ptr)
return beliefs32.astype('float')
def get_node_expectedutils(self, node):
utils32 = self.ntc.getnodeexpectedutils(node.ptr)
return utils32.astype('float')
def enter_finding(self, node, evidence):
if isinstance(evidence, int):
stateindx = evidence
elif isinstance(evidence, basestring):
lowercaseStates = np.array([name.lower() for name in node.statenames])
stateindx = np.where(lowercaseStates == evidence.lower())[0][0]
stateindx = c_int(stateindx)
self.ntc.enterfinding(node.ptr, stateindx)
def retract_nodefindings(self, node):
self.ntc.retractnodefindings(node.ptr)
def retract_netfindings(self):
self.ntc.retractnetfindings(self.net)
def save_net(self, filename):
self.ntc.savenet(self.env, self.net, filename)
def set_node_kind(self, node, nodekind=NATURE_NODE):
self.ntc.set_node_kind(node, nodekind)
def get_node_funcstate(self, node, parentstate):
return self.ntc.getnodefuncstate(node.ptr, parentstate)
def find_nodenamed(self, nodename):
"""
find node by node name
"""
found = False
for i,node in enumerate(self.nodes):
if node.name == nodename:
found =True
break
if found:
return node
else:
return 'error'
nodeC1_R1 = ntc.newnode('RelativeDamage_Houses', 0, net_p)
ntc.setnodelevels(nodeC1_R1, 4,
np.asarray([0, 0, 23.5, 47, 50], dtype='float64'))
ntc.setnodetitle(nodeC1_R1, 'Houses - Relative Damage (%)')
# define links
ntc.addlink(parent=nodeBC1, child=nodeH1_R1)
ntc.addlink(parent=nodeBC2, child=nodeH1_R1)
ntc.addlink(parent=nodeR1, child=nodeH1_R1)
ntc.addlink(parent=nodeH1_R1, child=nodeC1_R1)
ntc.setnodeprobs(nodeH1_R1, np.asarray([0, 0, 0], dtype='int'),
np.asarray([16.7, 44.7, 22.6, 16.1], dtype='float32'))
# obtain node list
#nl_p = ntc.getnetnodes(net_p)
# train with cas file
# ntc.revisecptsbycasefile(filename='BNcases.cas', nl_p=nl_p, updating=0, degree=1)
# compile the net
ntc.compilenet(net_p)
ntc.getnodeprobs(nodeH1_R1,
np.zeros(20,
dtype='int')) #np.asarray([0, 0, 0], dtype='int'))
# enable auto updating
ntc.setautoupdate(net_p)
# save net
ntc.savenet(env, net_p, 'BNZeebrugge.dne')
ntc.compilenet(net_p)
# enable auto updating
ntc.setautoupdate(net_p)
# prior belief
beliefs = ntc.getnodebeliefs(Tuberculosis)
belief = beliefs[0]
print 'The probability of tuberculosis is {:f}\n'.format(belief)
# posterior belief
ntc.enterfinding (XRay, 0)
beliefs = ntc.getnodebeliefs(Tuberculosis)
belief = beliefs[0]
print 'Given an abnormal X-ray,'
print ' the probability of tuberculosis is {:f}\n'.format(belief)
ntc.enterfinding (VisitAsia, 0)
beliefs = ntc.getnodebeliefs(Tuberculosis)
belief = beliefs[0]
print 'Given an abnormal X-ray and a visit to Asia,'
print ' the probability of tuberculosis is {:f}\n'.format(belief)
ntc.enterfinding (Cancer, 0)
beliefs = ntc.getnodebeliefs(Tuberculosis)
belief = beliefs[0]
print 'Given abnormal X-ray, Asia visit, and lung cancer,'
print ' the probability of tuberculosis is {:f}\n'.format(belief)
# save net
ntc.savenet(env, net_p, 'ChestClinic.dne')
class Network(object):
def __init__(self, name, lic=None, file=None):
if file is None:
self.name = name
self.ntc = Netica()
self.env = self.ntc.newenv(lic)
self.ntc.initenv(self.env)
self.net = self.ntc.newnet(name, self.env)
self.nodes = []
else:
self.name = name
self.ntc = Netica()
self.env = self.ntc.newenv(lic)
self.ntc.initenv(self.env)
self.net = self.ntc.opennet(self.env, file)
nodelist_p = self.ntc.getnetnodes(self.net)
numnode = self.ntc.lengthnodelist(nodelist_p)
self.nodes = []
for i in range(numnode):
nodei_p = self.ntc.nthnode(nodelist_p, i)
nodename = self.ntc.getnodename(nodei_p)
statename = 'init'
statenames = []
istate = 0
while statename != 'error':
statename = self.ntc.getnodestatename(nodei_p, istate)
statenames.append(statename)
istate += 1
statenames = statenames[:-1]
# default: no parents and continuous: therefore not the original nodes
nodei = Node(nodename, parents=[], rvname='continuous')
nodei.set_node_ptr(nodei_p)
nodei.set_node_state_name(statenames)
self.nodes.append(nodei)
def add_nodes(self, nodes):
for node in nodes:
if (node.nodekind == NATURE_NODE) and (node.cpt is None or
(node.cpt.size == 0)):
print "assign {} cpt first before add to network".format(
node.name)
#nodeptr = self.ntc.newnode(node.name, node.cpt.shape[0], self.net)
#node.set_pointer(nodeptr)
#node.set_net(self)
node.add_to_net(self)
self.nodes.append(node)
def add_link(self):
for node in self.nodes:
if node.parents is not None:
for parentNode in node.parents:
self.ntc.addlink(parent=parentNode.ptr, child=node.ptr)
def define_nodes(self):
"""define nodes can only be implemented after all links are added
"""
for node in self.nodes:
node.define()
def compile_net(self):
self.ntc.compilenet(self.net)
def set_autoupdate(self):
self.ntc.setautoupdate(self.net)
def get_node_beliefs(self, node):
beliefs32 = self.ntc.getnodebeliefs(node.ptr)
return beliefs32.astype('float')
def get_node_expectedutils(self, node):
utils32 = self.ntc.getnodeexpectedutils(node.ptr)
return utils32.astype('float')
def enter_finding(self, node, evidence):
if isinstance(evidence, int):
stateindx = evidence
elif isinstance(evidence, basestring):
lowercaseStates = np.array(
[name.lower() for name in node.statenames])
stateindx = np.where(lowercaseStates == evidence.lower())[0][0]
stateindx = c_int(stateindx)
self.ntc.enterfinding(node.ptr, stateindx)
def retract_nodefindings(self, node):
self.ntc.retractnodefindings(node.ptr)
def retract_netfindings(self):
self.ntc.retractnetfindings(self.net)
def save_net(self, filename):
self.ntc.savenet(self.env, self.net, filename)
def set_node_kind(self, node, nodekind=NATURE_NODE):
self.ntc.set_node_kind(node, nodekind)
def get_node_funcstate(self, node, parentstate):
return self.ntc.getnodefuncstate(node.ptr, parentstate)
def find_nodenamed(self, nodename):
"""
find node by node name
"""
found = False
for i, node in enumerate(self.nodes):
if node.name == nodename:
found = True
break
if found:
return node
else:
return 'error'
ntc.compilenet(net_p)
# enable auto updating
ntc.setautoupdate(net_p)
# prior belief
beliefs = ntc.getnodebeliefs(Tuberculosis)
belief = beliefs[0]
print 'The probability of tuberculosis is {:f}\n'.format(belief)
# posterior belief
ntc.enterfinding(XRay, 0)
beliefs = ntc.getnodebeliefs(Tuberculosis)
belief = beliefs[0]
print 'Given an abnormal X-ray,'
print ' the probability of tuberculosis is {:f}\n'.format(belief)
ntc.enterfinding(VisitAsia, 0)
beliefs = ntc.getnodebeliefs(Tuberculosis)
belief = beliefs[0]
print 'Given an abnormal X-ray and a visit to Asia,'
print ' the probability of tuberculosis is {:f}\n'.format(belief)
ntc.enterfinding(Cancer, 0)
beliefs = ntc.getnodebeliefs(Tuberculosis)
belief = beliefs[0]
print 'Given abnormal X-ray, Asia visit, and lung cancer,'
print ' the probability of tuberculosis is {:f}\n'.format(belief)
# save net
ntc.savenet(env, net_p, 'ChestClinic.dne')
nodeC1_R1 = ntc.newnode('RelativeDamage_Houses', 0, net_p)
ntc.setnodelevels(nodeC1_R1, 4, np.asarray([0, 0, 23.5, 47, 50], dtype='float64'))
ntc.setnodetitle(nodeC1_R1,'Houses - Relative Damage (%)')
# define links
ntc.addlink(parent=nodeBC1, child=nodeH1_R1)
ntc.addlink(parent=nodeBC2, child=nodeH1_R1)
ntc.addlink(parent=nodeR1, child=nodeH1_R1)
ntc.addlink(parent=nodeH1_R1, child=nodeC1_R1)
ntc.setnodeprobs(nodeH1_R1, np.asarray([0, 0, 0], dtype='int'), np.asarray([16.7,44.7,22.6,16.1],dtype='float32'))
# obtain node list
#nl_p = ntc.getnetnodes(net_p)
# train with cas file
# ntc.revisecptsbycasefile(filename='BNcases.cas', nl_p=nl_p, updating=0, degree=1)
# compile the net
ntc.compilenet(net_p)
ntc.getnodeprobs(nodeH1_R1, np.zeros(20, dtype='int')) #np.asarray([0, 0, 0], dtype='int'))
# enable auto updating
ntc.setautoupdate(net_p)
# save net
ntc.savenet(env, net_p, 'BNZeebrugge.dne')