"""Hyperparameters"""
LOG_DIR = "/Users/nhanitvn/Personal/Learning/DeepLearning/AE_ts/log_tb"
config = {} # Put all configuration information into the dict
config['num_layers'] = 2 # number of layers of stacked RNN's
config['hidden_size'] = 90 # memory cells in a layer
config['max_grad_norm'] = 5 # maximum gradient norm during training
config['batch_size'] = batch_size = 64
config['learning_rate'] = .005
config['crd'] = 1 # Hyperparameter for future generalization
config['num_l'] = 20 # number of units in the latent space
plot_every = 10 # after _plot_every_ GD steps, there's console output
max_iterations = 1000 # maximum number of iterations
dropout = 0.8 # Dropout rate
"""Load the data"""
X_train, X_val, y_train, y_val = open_data(direc)
N = X_train.shape[0]
Nval = X_val.shape[0]
D = X_train.shape[1]
config['sl'] = sl = D # sequence length
print('We have %s observations with %s dimensions' % (N, D))
# Organize the classes
num_classes = len(np.unique(y_train))
base = np.min(y_train) # Check if data is 0-based
if base != 0:
y_train -= base
y_val -= base
# Plot data # and save high quality plt.savefig('data_examples.eps', format='eps', dpi=1000)
config = dict() # Put all configuration information into the dict
config['num_layers'] = 2 # number of layers of stacked RNN's
config['hidden_size'] = 90 # memory cells in a layer
config['max_grad_norm'] = 5 # maximum gradient norm during training
config['batch_size'] = batch_size = 64
config['learning_rate'] = .005
config['crd'] = 1 # Hyperparameter for future generalization
config['num_l'] = 20 # number of units in the latent space
plot_every = 100 # after _plot_every_ GD steps, there's console output
max_iterations = 1000 # maximum number of iterations
dropout = 0.8 # Dropout rate
# Load the data
X_train, X_val, y_train, y_val = open_data(
'/home/rob/Dropbox/ml_projects/LSTM/UCR_TS_Archive_2015')
N = X_train.shape[0]
Nval = X_val.shape[0]
D = X_train.shape[1]
config['sl'] = sl = D # sequence length
print('We have %s observations with %s dimensions' % (N, D))
# Organize the classes
num_classes = len(np.unique(y_train))
base = np.min(y_train) # Check if data is 0-based
if base != 0:
y_train -= base
y_val -= base
# Plot data # and save high quality plt.savefig('data_examples.eps', format='eps', dpi=1000)
direc = './'
LOG_DIR = './'
config = {} # Put all configuration information into the dict
config['num_layers'] = 2 # number of layers of stacked RNN's
config['hidden_size'] = 90 # memory cells in a layer
config['max_grad_norm'] = 5 # maximum gradient norm during training
config['batch_size'] = batch_size = 64
config['learning_rate'] = .005
config['crd'] = 1 # Hyperparameter for future generalization
config['num_l'] = 1 # number of units in the latent space
plot_every = 100 # after _plot_every_ GD steps, there's console output
max_iterations = 10000 # maximum number of iterations
dropout = 0.8 # Dropout rate
"""Load the data"""
X_train, X_val, y_train, y_val = open_data('./UCR_TS_Archive_2015')
X_train, X_train_out = X_train[:, :-1], X_train[:, -1]
X_val, X_val_out = X_val[:, :-1], X_val[:, -1]
N = X_train.shape[0]
Nval = X_val.shape[0]
D = X_train.shape[1]
config['sl'] = sl = D # sequence length
print('We have %s observations with %s dimensions' % (N, D))
# Organize the classes
num_classes = len(np.unique(y_train))
base = np.min(y_train) # Check if data is 0-based
if base != 0:
y_train -= base
y_val -= base
LOG_DIR = "./Saved_Model" # Directory for the logging
config = dict() # Put all configuration information into the dict
config['num_layers'] = 2 # number of layers of stacked RNN's
config['hidden_size'] = 90 # memory cells in a layer
config['max_grad_norm'] = 0.5 # maximum gradient norm during training
config['batch_size'] = batch_size = 64
config['learning_rate'] = .005
config['num_l'] = 20 # number of units in the latent space
plot_every = 100 # after _plot_every_ GD steps, there's console output
max_iterations = 100 # maximum number of iterations
dropout = 0.8 # Dropout rate
# Load the data
X_train, X_val = open_data('./Data/')
N = X_train.shape[0] # nbr of element in train db
Nval = X_val.shape[0] # nbr of element in val db
D = X_train.shape[1] # nbr of columns in train db
config['sl'] = sl = D # sequence length
print('We have %s observations with %s dimensions' % (N, D))
# Organize the classes
num_classes = 15
"""Training time!"""
model = Model(config)
sess = tf.Session()
perf_collect = np.zeros(
(2, int(np.floor(max_iterations / plot_every)))
) # np.floor -> a = np.array([-1.7,2.0]) np.floor(a) array([-2., 2.]) *** np.zeros Return a new array of given shape filled with zeros.