def run(llt=1.0):
h0 = CCGLexicon(make_hypothesis, words=all_words, alpha=0.9, palpha=0.9, likelihood_temperature=llt)
fbs = FiniteBestSet(N=10)
from LOTlib.Inference.MetropolisHastings import mh_sample
for h in lot_iter(mh_sample(h0, data, SAMPLES)):
fbs.add(h, h.posterior_score)
return fbs
def run(*args):
# starting hypothesis -- here this generates at random
h0 = GaussianLOTHypothesis(grammar,
prior_temperature=PRIOR_TEMPERATURE)
# We store the top 100 from each run
pq = FiniteBestSet(100, max=True, key="posterior_score")
pq.add(mh_sample(h0, data, STEPS, skip=SKIP))
return pq
def run(data_size):
print "Running ", data_size
# We store the top 100 from each run
hypset = FiniteBestSet(TOP_COUNT, max=True)
# initialize the data
data = generate_data(data_size)
# starting hypothesis -- here this generates at random
learner = GriceanQuantifierLexicon(make_my_hypothesis, my_weight_function)
# We will defautly generate from null the grammar if no value is specified
for w in target.all_words():
learner.set_word(w)
# populate the finite sample by running the sampler for this many steps
for x in mh_sample(learner, data, SAMPLES, skip=0):
hypset.push(x, x.posterior_score)
return hypset
from LOTlib.Hypotheses.LOTHypothesis import LOTHypothesis
from LOTlib.Inference.Samplers.MetropolisHastings import mh_sample
from LOTlib.Examples.Quantifier.Model import *
ALPHA = 0.9
SAMPLES = 100000
DATA_SIZE = 1000
if __name__ == "__main__":
## sample the target data
data = generate_data(DATA_SIZE)
W = 'every'
# Now to use it as a LOTHypothesis, we need data to have an "output" field which is true/false for whether its the target word. This is then used by LOTHypothesis.compute_likelihood to see if we match or not with whether a word was said (ignoring the other words -- that's why its a pseudolikelihood)
for di in data:
di.output = (di.word == W)
#print (di.word == W)
FBS = FiniteBestSet(max=True, N=100)
H = LOTHypothesis(grammar, args=['A', 'B', 'S'], ALPHA=ALPHA)
# Now just run the sampler with a LOTHypothesis
for s in mh_sample(H, data, SAMPLES, skip=10):
#print s.lp, "\t", s.prior, "\t", s.likelihood, "\n", s, "\n\n"
FBS.push(s, s.lp)
for k in reversed(FBS.get_all(sorted=True)):
print k.lp, k.prior, k.likelihood, k
learner = GriceanQuantifierLexicon(make_my_hypothesis, my_weight_function)
# We will defautly generate from null the grammar if no value is specified
for w in target.all_words():
learner.set_word(w)
# populate the finite sample by running the sampler for this many steps
for x in mh_sample(learner, data, SAMPLES, skip=0):
hypset.push(x, x.posterior_score)
return hypset
if __name__ == "__main__":
# # # # # # # # # # # # # # # # # # # # # # # # # # # # # # #
# MPI interface
# Map. SimpleMPI will use a normal MAP if we are not running in MPI
allret = MPI_map(run, map(lambda x: [x],
DATA_AMOUNTS * CHAINS)) # this many chains
## combine into a single hypothesis set and save
outhyp = FiniteBestSet(max=True)
for r in allret:
print "# Merging ", len(r)
outhyp.merge(r)
import pickle
pickle.dump(outhyp, open(OUT_PATH, 'w'))
def run(*args):
#print "# Running data"
global hypotheses
data_size = args[0]
p_representation = defaultdict(
int) # how often do you get the right representation
p_response = defaultdict(int) # how often do you get the right response?
p_representation_literal = defaultdict(
int) # how often do you get the right representation
p_response_literal = defaultdict(
int) # how often do you get the right response?
p_representation_presup = defaultdict(
int) # how often do you get the right representation
p_response_presup = defaultdict(
int) # how often do you get the right response?
#print "# Generating data"
data = generate_data(data_size)
# recompute these
#print "# Computing posterior"
#[ x.unclear_functions() for x in hypotheses ]
[x.compute_posterior(data) for x in hypotheses]
# normalize the posterior in fs
#print "# Computing normalizer"
Z = logsumexp([x.posterior_score for x in hypotheses])
# and output the top hypotheses
qq = FiniteBestSet(max=True, N=25)
for h in hypotheses:
qq.push(h, h.posterior_score) # get the tops
for i, h in enumerate(qq.get_all(sorted=True)):
for w in h.all_words():
fprintn(8,
data_size,
i,
w,
h.posterior_score,
q(h.value[w]),
f=options.OUT_PATH + "-hypotheses." + str(get_rank()) +
".txt")
# and compute the probability of being correct
#print "# Computing correct probability"
for h in hypotheses:
hstr = str(h)
#print data_size, len(data), exp(h.posterior_score), correct[ str(h)+":"+w ]
for w in words:
p = exp(h.posterior_score - Z)
key = w + ":" + hstr
p_representation[w] += p * (agree_pct[key] == 1.)
p_representation_presup[w] += p * (
agree_pct_presup[key] == 1.
) # if we always agree with the target, then we count as the right rep.
p_representation_literal[w] += p * (agree_pct_literal[key] == 1.)
# and just how often does the hypothesis agree?
p_response[w] += p * agree_pct[key]
p_response_presup[w] += p * agree_pct_presup[key]
p_response_literal[w] += p * agree_pct_literal[key]
#print "# Outputting"
for w in words:
fprintn(10,
str(get_rank()),
q(w),
data_size,
p_representation[w],
p_representation_presup[w],
p_representation_literal[w],
p_response[w],
p_response_presup[w],
p_response_literal[w],
f=options.OUT_PATH + "-stats." + str(get_rank()) + ".txt")
return 0
def make_h0(value=None):
return GaussianLOTHypothesis(grammar, value=value)
if __name__ == "__main__":
# # # # # # # # # # # # # # # # # # # # # # # # # # # #
# the running function
def run(*args):
# starting hypothesis -- here this generates at random
h0 = GaussianLOTHypothesis(grammar,
prior_temperature=PRIOR_TEMPERATURE)
# We store the top 100 from each run
pq = FiniteBestSet(100, max=True, key="posterior_score")
pq.add(mh_sample(h0, data, STEPS, skip=SKIP))
return pq
finitesample = FiniteBestSet(max=True) # the finite sample of all
results = map(run, [[None]] * CHAINS) # Run on a single core
finitesample.merge(results)
## and display
for r in finitesample.get_all(decreasing=False, sorted=True):
print r.posterior_score, r.prior, r.likelihood, qq(str(r))
fbs.add(h, h.posterior_score)
return fbs
## ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
### MPI map
## ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
from SimpleMPI.MPI_map import MPI_map, is_master_process
allret = MPI_map(run, map(lambda x: [x], [0.01, 0.1, 1.0] * 100 ))
if is_master_process():
allfbs = FiniteBestSet(max=True)
allfbs.merge(allret)
H = allfbs.get_all()
for h in H:
h.likelihood_temperature = 0.01 # on what set of data we want?
h.compute_posterior(data)
# show the *average* ll for each hypothesis
for h in sorted(H, key=lambda h: h.posterior_score):
print h.posterior_score, h.prior, h.likelihood, h.likelihood_temperature
print h
# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
## Play around with some different inference schemes