from model import *
import config as cfg
import time
import os
loss, train_decode_result, pred_decode_result = build_network(is_training=True)
optimizer = tf.train.MomentumOptimizer(learning_rate=cfg.learning_rate,
momentum=cfg.momentum,
use_nesterov=True)
train_op = optimizer.minimize(loss)
saver = tf.train.Saver(max_to_keep=5)
img, label = cfg.read_data(cfg.train_dir)
num_train_samples = img.shape[0]
num_batches_per_epoch = int(num_train_samples / cfg.BATCH_SIZE)
sess = tf.Session()
sess.run(tf.global_variables_initializer())
with tf.name_scope('summaries'):
tf.summary.scalar("cost", loss)
summary_op = tf.summary.merge_all()
writer = tf.summary.FileWriter(cfg.LOGS_PATH)
target_in, target_out = cfg.label2int(label)
for cur_epoch in range(cfg.EPOCH):
shuffle_idx = np.random.permutation(num_train_samples)
train_cost = 0
start_time = time.time()
batch_time = time.time()
# the tracing part
for cur_batch in range(num_batches_per_epoch):
batch_time = time.time()
indexs = [
saver = tf.train.Saver(max_to_keep=5)
sess = tf.Session()
sess.run(tf.global_variables_initializer())
with tf.name_scope('summaries'):
tf.summary.scalar("cost", loss)
summary_op = tf.summary.merge_all()
writer = tf.summary.FileWriter(cfg.LOGS_PATH)
if cfg.is_restore:
ckpt = tf.train.latest_checkpoint(cfg.CKPT_DIR)
if ckpt:
saver.restore(sess,ckpt)
print('restore from the checkpoint{0}'.format(ckpt))
img,label=cfg.read_data(cfg.train_dir)
val_img,val_label=cfg.read_data(cfg.val_dir)
num_train_samples=img.shape[0]
num_batches_per_epoch = int(num_train_samples/cfg.BATCH_SIZE)
target_in,target_out=cfg.label2int(label)
for cur_epoch in range(cfg.EPOCH):
shuffle_idx = np.random.permutation(num_train_samples)
train_cost = 0
start_time = time.time()
batch_time = time.time()
# the tracing part
for cur_batch in range(num_batches_per_epoch):
val_img,val_label=shuffle(val_img,val_label)
batch_time = time.time()
indexs = [shuffle_idx[i % num_train_samples] for i in
d_rmses.append(d_rmse)
omegas.append(omega)
# break
if path.isfile(outfile):
print('%s exists' % path.basename(outfile))
return
data = {
'e': e_rmses,
'd': d_rmses,
'o': omegas,
}
config.make_file_dir(outfile)
pickle.dump(data, open(outfile, 'wb'))
#### load data
n_users, n_items, n_rates, indexes = config.read_data(song_file)
cmpl_rates = config.complete_rate(indexes)
dataset = n_users, n_items, n_rates, indexes, cmpl_rates
recom_list = config.provide_recom(indexes, cmpl_rates)
alpha = f_alpha
# v_omegas = np.arange(0.5, 1.55, 0.1)
v_omegas = np.arange(1.0, -0.05, -0.1)
v_omegas = np.arange(0.80, 1.25, 0.05)
risk = 'mae', np.absolute
n_mcar = 500
vary_error(n_mcar, dataset, recom_list, risk)
exit()
n_mcar = 50
vary_error(n_mcar, dataset, recom_list, risk)
from model import *
import config as cfg
import time
import os
loss,train_decode_result,pred_decode_result=build_network(is_training=True)
saver = tf.train.Saver()
sess = tf.Session()
ckpt = tf.train.latest_checkpoint(cfg.CKPT_DIR)
if ckpt:
saver.restore(sess,ckpt)
print('restore from the checkpoint{0}'.format(ckpt))
else:
print('failed to load ckpt')
val_img,_=cfg.read_data(cfg.val_dir)
val_predict = sess.run(pred_decode_result,feed_dict={image: val_img})
predit = cfg.int2label(np.argmax(val_predict, axis=2))
print(predit)