def train(self, iterations=10, maxmiss=10):
"""
Train using Recommended Greedy Algorithm
"""
scores = {
'Iteration': [],
#'Network':[],
'Score': []
}
score_check = maxmiss
niter = iterations
nodes = [i for i in self.data.columns]
best = score_pom(export_pom(self.net, by='label'), self.data)
#print("START LOOP")
while score_check > 0 and niter > 0:
n = dc(self.net)
upd = set()
ops = [n.add_edge, n.del_edge, n.rev_edge]
for f in ops:
edge = np.random.choice(nodes, size=2, replace=False)
f(edge[0], edge[1])
upd.add(edge[0])
upd.add(edge[1])
if n.acyclic():
n.calc_cpt(self.data, alpha=self.alpha, change=upd)
score = score_pom(export_pom(n, by='label'), self.data)
scores['Iteration'].append(iterations - niter)
#scores['Network'].append(n)
scores['Score'].append(score)
#print(best, score, niter, score_check)
if score > best:
self.net = n
best = score
niter = niter - 1
score_check = maxmiss
continue
else:
score_check = score_check - 1
niter = niter - 1
continue
else:
niter = niter - 1
continue
self.scores = scores
def specificity(net, data):
res = {}
m = export_pom(net, by=net.by)
for c in data.columns:
#print(c)
res[c] = sp(pred_bin(m, data, c))
return res
def sensetivity(net, data):
res = {}
m = export_pom(net, by=net.by)
for c in data.columns:
#print(pred_bin(m, data, c))
res[c] = sn(pred_bin(m, data, c))
return res
def accuracy(net, data):
res = {}
m = export_pom(net, by=net.by)
for c in data.columns:
#print(c)
res[c] = ac(pred_mult(m, data, c))
return res
def train(self, iterations=10, maxmiss=10):
"""
Train using Recommended Greedy Algorithm
"""
scores = {'Iteration': [], 'Network': [], 'Score': []}
score_check = maxmiss
niter = iterations
nodes = [i for i in self.data.columns]
best = score_pom(export_pom(self.net, by='label'), self.data)
#print("START LOOP")
while score_check > 0 and niter > 0:
n = net(data=self.data)
n.import_dag(self.net.export_dag())
ops = [n.add_edge, n.del_edge, n.rev_edge]
for f in ops:
# Choose the first node in a uniform, random way
v1 = np.random.choice(nodes)
# Choose the second with probabilities weighted by mi
v2 = np.random.choice(self.mi_weights[v1].index,
p=self.mi_weights[v1])
f(v1, v2)
if n.acyclic():
n.calc_cpt(self.data, alpha=self.alpha)
score = score_pom(export_pom(n, by='label'), self.data)
scores['Iteration'].append(iterations - niter)
scores['Network'].append(n)
scores['Score'].append(score)
#print(best, score, niter, score_check)
if score > best:
self.net = n
best = score
niter = niter - 1
score_check = maxmiss
continue
else:
score_check = score_check - 1
niter = niter - 1
continue
else:
niter = niter - 1
continue
self.scores = scores
t_res = t_res.append(
pd.DataFrame(grd.scores).assign(Trial=i, Learner='GREEDY', Net="ds1"))
t_res = t_res.append(
pd.DataFrame(cgm.scores).assign(Trial=i, Learner='CASGMM', Net="ds1"))
t_res = t_res.append(
pd.DataFrame(cmd.scores).assign(Trial=i, Learner='CASMOD', Net="ds1"))
t_res = t_res.append(
pd.DataFrame(cjn.scores).assign(Trial=i, Learner='CASJNK', Net="ds1"))
g_res = g_res.append(pd.DataFrame({
i[0]: [i[1]]
for i in edge_hits(export_pom(grd.net, by='label'),
export_pom(tn1, by='label')).items()
}).assign(Trial=i, Learner='GREEDY', Net="ds1"),
sort=False)
g_res = g_res.append(pd.DataFrame({
i[0]: [i[1]]
for i in edge_hits(export_pom(cgm.net, by='label'),
export_pom(tn1, by='label')).items()
}).assign(Trial=i, Learner='CASGMM', Net="ds1"),
sort=False)
g_res = g_res.append(pd.DataFrame({
i[0]: [i[1]]
for i in edge_hits(export_pom(cmd.net, by='label'),
export_pom(tn1, by='label')).items()
t_res=t_res.append(
pd.DataFrame(
cmd.scores).assign(Trial = i, Learner = 'CASMOD', Net="ds2")
)
t_res=t_res.append(
pd.DataFrame(
cjn.scores).assign(Trial = i, Learner = 'CASJNK', Net="ds2")
)
g_res = g_res.append(
pd.DataFrame(
{
i[0]:[i[1]]
for i in edge_hits(
export_pom(grd.net, by='label'),
export_pom(tn2, by='label')
).items()
}
).assign(Trial = i, Learner = 'GREEDY', Net="ds2"),
sort=False
)
g_res = g_res.append(
pd.DataFrame(
{
i[0]:[i[1]]
for i in edge_hits(
export_pom(cgm.net, by='label'),
export_pom(tn2, by='label')
tn2 = net(data=ds2)
for i in range(0, 4):
tn2.add_edge('G' + str(i), 'G' + str(i + 1))
for i in range(5, 9):
tn2.add_edge('G' + str(i), 'G' + str(i + 1))
for i in range(10, 14):
tn2.add_edge('G' + str(i), 'G' + str(i + 1))
for i in range(15, 19):
tn2.add_edge('G' + str(i), 'G' + str(i + 1))
tn2.calc_cpt(ds2, alpha=0.00001)
export_pom(tn1, by='label')
# Data Set 3: 4 Variables generated by sampling from a fixed set of conditional
# probability tables (Example from: Finsen Jenn)
data = pd.DataFrame({
'Icy': [0, 1],
'Holmes': [0, 1],
'Watson': [0, 1],
'Ambulance': [0, 1]
})
tn3 = net(data=data)
tn3.add_edge('Icy', 'Watson')
tn3.add_edge('Icy', 'Holmes')
tn3.add_edge('Holmes', 'Ambulance')
]
data['G4'] = [
pfun(data['G3'][i])
for i in range(0,len(data['G3']))
]
data = pd.DataFrame(data)
data = data[['G1', 'G2', 'G3', 'G4']]
a = net(data=data)
a.add_edge(1,0)
a.add_edge(2,3)
a.calc_cpt(data)
m = export_pom(a, by='label')
v = [i.name for i in m.states]
# Function that calculates the probability of a row given a set of nodes
def f(r, nds):
pr = []
for n in nds:
c=n.cpt.columns.drop('Prob')
pr.append(
#np.log(
n.cpt.set_index(list(c)).loc[tuple(r[c])].values[0]
# )
)
return(np.array(pr).prod())