def lstm_with_static_feature_model_experiments(result_file):
if ExperimentSetup.data_source == 'lu':
data_set = read_data_lu()
else:
data_set = read_data_sun()
static_feature = data_set.static_feature
dynamic_feature = data_set.dynamic_feature
labels = data_set.labels
static_n_features = static_feature.shape[1]
dynamic_n_features = dynamic_feature.shape[2]
time_steps = dynamic_feature.shape[1]
n_output = labels.shape[1]
model = LSTMWithStaticFeature(
static_n_features,
dynamic_n_features,
time_steps,
ExperimentSetup.lstm_size,
n_output,
batch_size=ExperimentSetup.batch_size,
optimizer=tf.train.AdamOptimizer(ExperimentSetup.learning_rate),
epochs=ExperimentSetup.epochs,
output_n_epochs=ExperimentSetup.output_n_epochs)
return model_experiments(model, data_set, result_file)
def resnet_model_experiments(result_file):
if ExperimentSetup.data_source == 'lu':
data_set = read_data_lu()
else:
data_set = read_data_sun()
static_feature = data_set.static_feature
labels = data_set.labels
static_n_features = static_feature.shape[1]
n_output = labels.shape[1]
model = ResNet(static_n_features,
n_output,
batch_size=ExperimentSetup.batch_size,
optimizer=tf.train.AdamOptimizer(
ExperimentSetup.learning_rate),
epochs=ExperimentSetup.epochs,
output_n_epochs=ExperimentSetup.output_n_epochs)
return model_experiments(model, data_set, result_file)
def bidirectional_lstm_model_experiments(result_file):
if ExperimentSetup.data_source == 'lu':
data_set = read_data_lu()
else:
data_set = read_data_sun()
dynamic_feature = data_set.dynamic_feature
labels = data_set.labels
num_features = dynamic_feature.shape[2]
time_steps = dynamic_feature.shape[1]
n_output = labels.shape[1]
model = BidirectionalLSTMModel(
num_features,
time_steps,
ExperimentSetup.lstm_size,
n_output,
batch_size=ExperimentSetup.batch_size,
optimizer=tf.train.AdamOptimizer(ExperimentSetup.learning_rate),
epochs=ExperimentSetup.epochs,
output_n_epoch=ExperimentSetup.output_n_epochs)
return model_experiments(model, data_set, result_file)