# choose to corrupt one part of the triple
to_mod = data_rng.choice(3)
# corrupt with some other part
if to_mod == 0:
while word_a_new == word_a:
word_a_new = data_rng.choice(relationships.indices_of_words_in_synsets) #sample_cumulative_discrete_distribution(ngram_reader.cumulative_word_frequencies, rng=data_rng)
elif to_mod == 1:
while word_b_new == word_b:
word_b_new = data_rng.choice(relationships.indices_of_words_in_synsets) #sample_cumulative_discrete_distribution(ngram_reader.cumulative_word_frequencies, rng=data_rng)
elif to_mod == 2:
while rel_index_new == rel_index:
# rel_index_new = data_rng.randint(N_relationships)
rel_index_new = data_rng.randint(relationships.N_relationships)
augmented_cost, cost = model.update_v(word_a, word_b, rel_index, word_a_new, word_b_new, rel_index_new)
if not np.isfinite(cost):
print 'nan detected'
save_model('nan_dump.pkl.gz')
import IPython
IPython.embed()
costs.append(cost)
augmented_costs.append(augmented_cost)
if i % print_freq == 0:
sys.stdout.write('\r k %i: pair : %d / %d' % (model.k, i, block_size))
sys.stdout.flush()
if args['semantic_blocks_to_run'] > 1:
print
print '%i intermediate mean %f' % (block_num, np.mean(costs[-block_size:]))
this_count = 0
augmented_costs = []
costs = []
for block_num in xrange(args['semantic_blocks_to_run']):
for i in xrange(args['semantic_block_size']):
train_i = -1
while train_i not in indices_in_intersection:
train_i = sample_cumulative_discrete_distribution(ngram_reader.cumulative_word_frequencies, rng=data_rng)
for j in xrange(args['k_nearest']):
train_j = -1
while train_j not in indices_in_intersection:
train_j = sample_cumulative_discrete_distribution(ngram_reader.cumulative_word_frequencies, rng=data_rng)
if word_similarity.word_pairwise_sims[train_i, train_j] == -np.inf:
continue
sim = word_similarity.word_pairwise_sims[train_i, train_j]
augmented_cost, cost = model.update_v(train_i, train_j, sim)
augmented_costs.append(augmented_cost)
costs.append(cost)
if i % print_freq == 0:
sys.stdout.write('\r k %i: pair : %d / %d' % (model.k, i, args['semantic_block_size']))
sys.stdout.flush()
if args['semantic_blocks_to_run'] > 1:
print
print '%i intermediate mean %f' % (block_num, np.mean(costs[-args['semantic_block_size']:]))
print
stats_for_k['semantic_mean'] = np.mean(costs)
stats_for_k['semantic_std'] = np.std(costs)
print 'semantic mean cost \t%f' % stats_for_k['semantic_mean']
print 'semantic std cost \t%f' % stats_for_k['semantic_std']
while train_i not in indices_in_intersection:
train_i = sample_cumulative_discrete_distribution(
ngram_reader.cumulative_word_frequencies,
rng=data_rng)
for j in xrange(args['k_nearest']):
train_j = -1
while train_j not in indices_in_intersection:
train_j = sample_cumulative_discrete_distribution(
ngram_reader.cumulative_word_frequencies,
rng=data_rng)
if word_similarity.word_pairwise_sims[
train_i, train_j] == -np.inf:
continue
sim = word_similarity.word_pairwise_sims[train_i,
train_j]
augmented_cost, cost = model.update_v(
train_i, train_j, sim)
augmented_costs.append(augmented_cost)
costs.append(cost)
if i % print_freq == 0:
sys.stdout.write(
'\r k %i: pair : %d / %d' %
(model.k, i, args['semantic_block_size']))
sys.stdout.flush()
if args['semantic_blocks_to_run'] > 1:
print
print '%i intermediate mean %f' % (
block_num, np.mean(
costs[-args['semantic_block_size']:]))
print
word_a_new = data_rng.choice(
relationships.indices_of_words_in_synsets
) #sample_cumulative_discrete_distribution(ngram_reader.cumulative_word_frequencies, rng=data_rng)
elif to_mod == 1:
while word_b_new == word_b:
word_b_new = data_rng.choice(
relationships.indices_of_words_in_synsets
) #sample_cumulative_discrete_distribution(ngram_reader.cumulative_word_frequencies, rng=data_rng)
elif to_mod == 2:
while rel_index_new == rel_index:
# rel_index_new = data_rng.randint(N_relationships)
rel_index_new = data_rng.randint(
relationships.N_relationships)
augmented_cost, cost = model.update_v(
word_a, word_b, rel_index, word_a_new, word_b_new,
rel_index_new)
if not np.isfinite(cost):
print 'nan detected'
save_model('nan_dump.pkl.gz')
import IPython
IPython.embed()
costs.append(cost)
augmented_costs.append(augmented_cost)
if i % print_freq == 0:
sys.stdout.write('\r k %i: pair : %d / %d' %
(model.k, i, block_size))
sys.stdout.flush()
if args['semantic_blocks_to_run'] > 1:
