labeled_sets = kfold_by_cat(labeled_docdict, labeled_size_experiment)
i = 0
for labeled_set in labeled_sets[:N_LABELED_SETS]:
i += 1
labeled_x, labeled_y = convert_docdict_to_array(labeled_set, vec, enc)
rowout = [labeled_x.shape[0], i]
nb = MultinomialNB(alpha=ALPHA)
nb.fit(labeled_x, labeled_y)
noem_score = nb.score(test_x, test_y)
experiment_scores_noem.append(noem_score)
rowout.append(noem_score)
em = NaiveBayesEM(unlabeled_x, ncats, alpha=ALPHA, labeled_x=labeled_x, labeled_y=labeled_y)
em.models.append(nb)
em.runEM()
nb_out = em.models[-1]
em_score = nb_out.score(test_x, test_y)
rowout.append(em_score)
this_em_iter_scores = []
for model in em.models:
this_em_iter_scores.append(model.score(test_x, test_y))
print "size = %s, score = %s, em_score = %s" % (labeled_size_experiment, noem_score, em_score)
with open(FNAME_OUT, "a") as fout:
for labeled_set in labeled_sets[:N_LABELED_SETS]:
i += 1
labeled_x, labeled_y = convert_docdict_to_array(
labeled_set, vec, enc)
rowout = [labeled_x.shape[0], alpha, i]
nb = MultinomialNB(alpha=alpha)
nb.fit(labeled_x, labeled_y)
noem_score = nb.score(test_x, test_y)
experiment_scores_noem.append(noem_score)
rowout.append(noem_score)
em = NaiveBayesEM(unlabeled_x,
ncats,
alpha=alpha,
labeled_x=labeled_x,
labeled_y=labeled_y)
em.models.append(nb)
em.runEM()
nb_out = em.models[-1]
em_score = nb_out.score(test_x, test_y)
rowout.append(em_score)
this_em_iter_scores = []
for model in em.models:
this_em_iter_scores.append(model.score(test_x, test_y))
print "size = %s, alpha = %s, score = %s, em_score = %s" % (
i = 0
total = len(labeled_sets)
for labeled_set in labeled_sets[:MAX_N_UNLABELED_SETS]:
print "testing size =", labeled_size_experiment, ", n =", i
i += 1
labeled_x, labeled_y = convert_docdict_to_array(labeled_set, vec, enc)
rowout = [labeled_x.shape[0], i]
nb = MultinomialNB(alpha=ALPHA)
nb.fit(labeled_x, labeled_y)
noem_score = nb.score(test_x, test_y)
experiment_scores_noem.append(noem_score)
rowout.append(noem_score)
em = NaiveBayesEM(unlabeled_x, ncats, alpha=ALPHA)
em.model = nb
em.runEM()
nb_out = em.model
em_score = nb_out.score(test_x, test_y)
experiment_scores_em.append(em_score)
rowout.append(em_score)
print rowout
with open(FNAME_OUT,'a') as fout:
fwriter = csv.writer(fout)
fwriter.writerow(rowout)