def main():
print("load initial model ...")
param_nn = cPickle.load(open(DIS_MODEL_FILE_NN))
assert param_nn is not None
discriminator = DIS(FEATURE_SIZE,
HIDDEN_SIZE,
D_WEIGHT_DECAY,
D_LEARNING_RATE,
loss='log',
param=param_nn)
generator = GEN(FEATURE_SIZE,
HIDDEN_SIZE,
G_WEIGHT_DECAY,
G_LEARNING_RATE,
param=param_nn)
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
sess = tf.Session(config=config)
sess.run(tf.initialize_all_variables())
print('start adversarial training')
p_best_val = 0.0
ndcg_best_val = 0.0
for epoch in range(30):
if epoch > 0:
# G generate negative for D, then train D
print('Training D ...')
generate_for_d(sess, generator, DIS_TRAIN_FILE)
train_size = ut.file_len(DIS_TRAIN_FILE)
for d_epoch in range(30):
index = 1
while True:
if index > train_size:
break
if index + BATCH_SIZE <= train_size + 1:
input_pos, input_neg = ut.get_batch_data(
DIS_TRAIN_FILE, index, BATCH_SIZE)
else:
input_pos, input_neg = ut.get_batch_data(
DIS_TRAIN_FILE, index, train_size - index + 1)
index += BATCH_SIZE
_ = sess.run(discriminator.d_updates,
feed_dict={
discriminator.pos_data: input_pos,
discriminator.neg_data: input_neg
})
p_5 = precision_at_k(sess,
discriminator,
query_pos_test,
query_pos_train,
query_url_feature,
k=5)
ndcg_5 = ndcg_at_k(sess,
discriminator,
query_pos_test,
query_pos_train,
query_url_feature,
k=5)
if p_5 > p_best_val:
p_best_val = p_5
ndcg_best_val = ndcg_5
discriminator.save_model(sess, GAN_MODEL_BEST_FILE)
print("Best: ", "dis p@5 ", p_5, "dis ndcg@5 ", ndcg_5)
elif p_5 == p_best_val:
if ndcg_5 > ndcg_best_val:
ndcg_best_val = ndcg_5
discriminator.save_model(sess, GAN_MODEL_BEST_FILE)
print("Best: ", "dis p@5 ", p_5, "dis ndcg@5 ", ndcg_5)
# Train G
print('Training G ...')
for g_epoch in range(50): # 50
for query in query_pos_train.keys():
pos_list = query_pos_train[query]
# candidate_list = list(set(query_url_feature[query].keys()) - set(pos_list))
candidate_list = list(query_url_feature[query].keys())
if len(candidate_list) <= 0:
continue
candidate_list_feature = [
query_url_feature[query][url] for url in candidate_list
]
candidate_list_feature = np.asarray(candidate_list_feature)
candidate_list_score = sess.run(
generator.pred_score,
{generator.pred_data: candidate_list_feature})
# softmax for all
exp_rating = np.exp(candidate_list_score)
prob = exp_rating / np.sum(exp_rating)
neg_index = np.random.choice(np.arange(len(candidate_list)),
size=[len(pos_list)],
p=prob)
neg_list = np.array(candidate_list)[neg_index]
pos_list_feature = [
query_url_feature[query][url] for url in pos_list
]
neg_list_feature = [
query_url_feature[query][url] for url in neg_list
]
neg_index = np.asarray(neg_index)
# every negative samples have a reward
neg_reward = sess.run(discriminator.reward,
feed_dict={
discriminator.pos_data:
pos_list_feature,
discriminator.neg_data:
neg_list_feature
})
# Method 1: softmax before gather
_ = sess.run(generator.gan_updates,
feed_dict={
generator.pred_data: candidate_list_feature,
generator.sample_index: neg_index,
generator.reward: neg_reward
})
print('Best p@5: ', p_best_val, 'Best ndcg@5: ', ndcg_best_val)
def main():
discriminator = DIS(FEATURE_SIZE,
HIDDEN_SIZE,
WEIGHT_DECAY,
D_LEARNING_RATE,
loss='log',
param=None)
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
sess = tf.Session(config=config)
sess.run(tf.initialize_all_variables())
print('start dynamic negative sampling with log ranking discriminator')
p_best_val = 0.0
ndcg_best_val = 0.0
for epoch in range(200):
generate_dns(sess, discriminator, DIS_TRAIN_FILE)
train_size = ut.file_len(DIS_TRAIN_FILE)
index = 1
while True:
if index > train_size:
break
if index + BATCH_SIZE <= train_size + 1:
input_pos, input_neg = ut.get_batch_data(
DIS_TRAIN_FILE, index, BATCH_SIZE)
else:
input_pos, input_neg = ut.get_batch_data(
DIS_TRAIN_FILE, index, train_size - index + 1)
index += BATCH_SIZE
input_pos = np.asarray(input_pos)
input_neg = np.asarray(input_neg)
_ = sess.run(discriminator.d_updates,
feed_dict={
discriminator.pos_data: input_pos,
discriminator.neg_data: input_neg
})
p_5 = precision_at_k(sess,
discriminator,
query_pos_test,
query_pos_train,
query_url_feature,
k=5)
ndcg_5 = ndcg_at_k(sess,
discriminator,
query_pos_test,
query_pos_train,
query_url_feature,
k=5)
if p_5 > p_best_val:
p_best_val = p_5
discriminator.save_model(sess, DNS_MODEL_BEST_FILE)
print("Best: ", " p@5 ", p_5, "ndcg@5 ", ndcg_5)
elif p_5 == p_best_val:
if ndcg_5 > ndcg_best_val:
ndcg_best_val = ndcg_5
discriminator.save_model(sess, DNS_MODEL_BEST_FILE)
print("Best: ", " p@5 ", p_5, "ndcg@5 ", ndcg_5)
sess.close()
param_best = cPickle.load(open(DNS_MODEL_BEST_FILE))
assert param_best is not None
discriminator_best = DIS(FEATURE_SIZE,
HIDDEN_SIZE,
WEIGHT_DECAY,
D_LEARNING_RATE,
loss='log',
param=param_best)
sess = tf.Session(config=config)
sess.run(tf.initialize_all_variables())
p_1_best = precision_at_k(sess,
discriminator_best,
query_pos_test,
query_pos_train,
query_url_feature,
k=1)
p_3_best = precision_at_k(sess,
discriminator_best,
query_pos_test,
query_pos_train,
query_url_feature,
k=3)
p_5_best = precision_at_k(sess,
discriminator_best,
query_pos_test,
query_pos_train,
query_url_feature,
k=5)
p_10_best = precision_at_k(sess,
discriminator_best,
query_pos_test,
query_pos_train,
query_url_feature,
k=10)
ndcg_1_best = ndcg_at_k(sess,
discriminator_best,
query_pos_test,
query_pos_train,
query_url_feature,
k=1)
ndcg_3_best = ndcg_at_k(sess,
discriminator_best,
query_pos_test,
query_pos_train,
query_url_feature,
k=3)
ndcg_5_best = ndcg_at_k(sess,
discriminator_best,
query_pos_test,
query_pos_train,
query_url_feature,
k=5)
ndcg_10_best = ndcg_at_k(sess,
discriminator_best,
query_pos_test,
query_pos_train,
query_url_feature,
k=10)
map_best = MAP(sess, discriminator_best, query_pos_test, query_pos_train,
query_url_feature)
mrr_best = MRR(sess, discriminator_best, query_pos_test, query_pos_train,
query_url_feature)
print("Best ", "p@1 ", p_1_best, "p@3 ", p_3_best, "p@5 ", p_5_best,
"p@10 ", p_10_best)
print("Best ", "ndcg@1 ", ndcg_1_best, "ndcg@3 ", ndcg_3_best, "ndcg@5 ",
ndcg_5_best, "p@10 ", ndcg_10_best)
print("Best MAP ", map_best)
print("Best MRR ", mrr_best)
D_WEIGHT_DECAY = 0.001
D_LEARNING_RATE = 0.0001
workdir = 'MQ2008-semi'
GAN_PAIRWISE_MODEL_BEST_FILE = workdir + '/gan/gan_best_nn.model'
query_url_feature =\
ut.load_all_query_url_feature(workdir + '/Large_norm.txt', FEATURE_SIZE)
query_pos_train = ut.get_query_pos(workdir + '/train.txt')
query_pos_test = ut.get_query_pos(workdir + '/test.txt')
param_best = cPickle.load(open(GAN_PAIRWISE_MODEL_BEST_FILE))
assert param_best is not None
discriminator_best = DIS(FEATURE_SIZE,
HIDDEN_SIZE,
D_WEIGHT_DECAY,
D_LEARNING_RATE,
loss='log',
param=param_best)
config = tf.ConfigProto()
config.gpu_options.allow_growth = True
sess = tf.Session(config=config)
sess.run(tf.initialize_all_variables())
p_1_best = precision_at_k(sess,
discriminator_best,
query_pos_test,
query_pos_train,
query_url_feature,
k=1)
p_3_best = precision_at_k(sess,