method = FLAGS.method trainset = melt.load_dataset(trainset_file) print "finish loading train set ",trainset_file testset = melt.load_dataset(testset_file) print "finish loading test set ", testset_file assert(trainset.num_features == testset.num_features) num_features = trainset.num_features print 'num_features: ', num_features print 'trainSet size: ', trainset.num_instances() print 'testSet size: ', testset.num_instances() print 'batch_size:', batch_size, ' learning_rate:', learning_rate, ' num_epochs:', num_epochs trainer = melt.gen_binary_classification_trainer(trainset) class LogisticRegresssion: def model(self, X, w): return melt.matmul(X,w) def forward(self, trainer): w = melt.init_weights([trainer.num_features, 1]) py_x = self.model(trainer.X, w) return py_x class Mlp: def model(self, X, w_h, w_o): h = tf.nn.sigmoid(melt.matmul(X, w_h)) # this is a basic mlp, think 2 stacked logistic regressions return tf.matmul(h, w_o) # note that we dont take the softmax at the end because our cost fn does that for us
method = FLAGS.method
trainset = melt.load_dataset(trainset_file)
print "finish loading train set ", trainset_file
testset = melt.load_dataset(testset_file)
print "finish loading test set ", testset_file
assert (trainset.num_features == testset.num_features)
num_features = trainset.num_features
print 'num_features: ', num_features
print 'trainSet size: ', trainset.num_instances()
print 'testSet size: ', testset.num_instances()
print 'batch_size:', batch_size, ' learning_rate:', learning_rate, ' num_epochs:', num_epochs
trainer = melt.gen_binary_classification_trainer(trainset)
class LogisticRegresssion:
def model(self, X, w):
return melt.matmul(X, w)
def forward(self, trainer):
w = melt.init_weights([trainer.num_features, 1])
py_x = self.model(trainer.X, w)
return py_x
class Mlp:
def model(self, X, w_h, w_o):
h = tf.nn.sigmoid(melt.matmul(
def train(self, trainset_file, testset_file, method, num_epochs,
learning_rate, model_path):
print 'batch_size:', batch_size, ' learning_rate:', learning_rate, ' num_epochs:', num_epochs
print 'method:', method
trainset = melt.load_dataset(trainset_file)
print "finish loading train set ", trainset_file
self.num_features = trainset.num_features
print 'num_features: ', self.num_features
print 'trainSet size: ', trainset.num_instances()
testset = melt.load_dataset(testset_file)
print "finish loading test set ", testset_file
assert (trainset.num_features == testset.num_features)
print 'testSet size: ', testset.num_instances()
algo = self.gen_algo(method)
trainer = melt.gen_binary_classification_trainer(trainset)
self.algo = algo
self.trainer = trainer
print 'trainer_type:', trainer.type
print 'trainer_index_only:', trainer.index_only
cost, train_op, predict_op, evaluate_op = self.foward(
algo, trainer, learning_rate)
#self.foward(algo, trainer, learning_rate)
config = None
if not FLAGS.show_device:
config = tf.ConfigProto()
else:
config = tf.ConfigProto(log_device_placement=True)
config.gpu_options.allocator_type = 'BFC'
self.session = tf.Session(config=config)
init = tf.initialize_all_variables()
self.session.run(init)
summary_writer = None
if FLAGS.use_summary:
tf.scalar_summary("cross_entropy", self.cost)
if FLAGS.use_auc_op:
tf.scalar_summary("auc", evaluate_op)
merged_summary_op = tf.merge_all_summaries()
summary_writer = tf.train.SummaryWriter(FLAGS.summary_path,
self.session.graph_def)
#os.system('rm -rf ' + FLAGS.model)
os.system('mkdir -p ' + FLAGS.model)
self.save_info(model_path)
for epoch in range(num_epochs):
if epoch > 0 and FLAGS.shuffle:
trainset = melt.load_dataset(trainset_file)
self.train_(trainset)
need_stop = self.test_(testset, epoch=epoch)
if need_stop:
print 'need stop as improve is smaller then %f' % FLAGS.min_improve
break
#print weight
#@FIXME
if epoch % FLAGS.save_epochs == 0 and not trainer.index_only:
self.save_model(model_path, epoch)
self.save_model(model_path)
if FLAGS.calibrate:
dataset = trainset
if not FLAGS.calibrate_trainset:
dataset = testset
self.calibrate_(dataset) #@TODO may be test set is right?
CalibratorFactory.Save(self.calibrator,
model_path + '/calibrator.bin')
#self.calibrator.Save(model_path + '/calibrator.bin')
self.calibrator.SaveText(model_path + '/calibrator.txt')
if FLAGS.use_summary:
teX, teY = testset.full_batch()
summary_str = self.session.run(merged_summary_op,
feed_dict=melt.gen_feed_dict(
self.trainer,
self.algo,
teX,
teY,
test_mode=True))
summary_writer.add_summary(summary_str, epoch)