print posteriorRestrictionToObjectsAndAdjectivesForSent([0, 1, 1], True, 0, 1)
print posteriorRestrictionToObjectsAndAdjectivesForSent([1, 0, 1], True, 1, 0)
def sigma(x):
return 1 / (1 + exp(-x))
alpha1 = 1.7
alpha2 = -0.3
u = 0
c = 0
logAvg = -0.6931471805599453
correct = .5
for epoch in range(200):
for point in prepareNgramData.dataset():
adj1 = point["AdjNum1"]
adj2 = point["AdjNum2"]
pmi1 = point["PMI1"]
pmi2 = point["PMI2"]
utilityLoss12 = posteriorRestrictionToObjectsAndAdjectivesForSent(
[adj1, adj2, 1], True, adj1, adj2)
utilityLoss21 = posteriorRestrictionToObjectsAndAdjectivesForSent(
[adj2, adj1, 1], True, adj2, adj1)
print[
utilityLoss12, utilityLoss21, "AGR", agreement[adj1],
agreement[adj2]
]
if (utilityLoss12 > utilityLoss21) != (agreement[adj1] <
agreement[adj2]):
print["ERROR", agreement[adj1], agreement[adj2]]
return sum(map(lambda x: x[0] * x[1], zip(surprisals, probabilities)))
# print "PROBABILITIES "+str(eventsForS1)
# return sum(map(lambda x:x * log(x), eventsForS1))
#print posteriorRestrictionToObjectsAndAdjectivesForSent([0,1,1], True, 0, 1)
#print posteriorRestrictionToObjectsAndAdjectivesForSent([1,0,1], True, 1, 0)
def sigma(x):
return 1 / (1 + exp(-x))
#dataset_size = sum(1 for _ in prepareNgramData.dataset())
dataiter = prepareNgramData.dataset()
dataset = [x for x in dataiter]
u = 0
c = 0
logAvg = -0.6931471805599453
correct = .5
#for epoch in range(200):
def guide(dataset):
# register the two variational parameters with Pyro
mu1 = pyro.param("mu1", Variable(torch.Tensor([(0.0)]),
requires_grad=True))
mu2 = pyro.param("mu2", Variable(torch.Tensor([(0.0)]),
requires_grad=True))