def start(self, x, y, **kwargs):
'''Start the scan. Note that you can use `Scan()` arguments as you
would otherwise directly interacting with `Scan()`.
`x` | array or list of arrays | prediction features
`y` | array or list of arrays | prediction outcome variable
`kwargs` | arguments | any `Scan()` argument can be passed here
'''
import talos
m = talos.autom8.AutoModel(self.task, self.experiment_name).model
try:
kwargs['params']
scan_object = talos.Scan(x,
y,
model=m,
experiment_name=self.experiment_name,
**kwargs)
except KeyError:
p = talos.autom8.AutoParams(task=self.task)
if self.max_param_values is not None:
p.resample_params(self.max_param_values)
params = p.params
scan_object = talos.Scan(x=x,
y=y,
params=params,
model=m,
experiment_name=self.experiment_name,
**kwargs)
return scan_object
def minimal():
x, y = ta.templates.datasets.iris()
p = {
'activation': ['relu', 'elu'],
'optimizer': ['Nadam', 'Adam'],
'losses': ['logcosh'],
'hidden_layers': [0, 1, 2],
'batch_size': [20, 30, 40],
'epochs': [10, 20]
}
def iris_model(x_train, y_train, x_val, y_val, params):
n_timesteps, n_features, n_outputs = trainX.shape[1], trainX.shape[
2], trainy.shape[1]
model = Sequential()
model.add(Dense(32, input_dim=4, activation=params['activation']))
model.add(Dense(3, activation='softmax'))
model.compile(optimizer=params['optimizer'], loss=params['losses'])
out = model.fit(x_train,
y_train,
verbose=0,
batch_size=params['batch_size'],
epochs=params['epochs'],
validation_data=[x_val, y_val])
return out, model
scan_object = ta.Scan(x, y, model=iris_model, params=p)
return scan_object
def NeuralNetScan(DataSet, Y):
seed = 7
test_size = 0.33
X_train, X_test, y_train, y_test = train_test_split(DataSet,
Y,
test_size=test_size,
random_state=seed)
sc = StandardScaler()
sc.fit(X_train)
X_train = sc.transform(X_train)
X_test = sc.transform(X_test)
p = {
'lr': (0.5, 5, 10),
'first_neuron': [4, 8, 16, 32, 64],
'hidden_layers': [0, 1, 2],
'epochs': [5],
'weight_regulizer': [None],
'emb_output_dims': [None],
'shape': ['brick'],
'optimizer': ['Adam'],
'losses': ['binary_crossentropy'],
'activation': ['relu', 'tanh'],
'last_activation': ['sigmoid']
}
t = ta.Scan(x=X_train,
y=y_train,
x_val=X_test,
y_val=y_test,
model=NeuralNetwork,
params=p,
experiment_name='Particle_DataSet')
BestParameters = t.data.sort_values(by='val_acc', ascending=False).iloc[0]
return t, BestParameters
def do_nn2_talos(x_train, x_test, y_train, y_test):
p = {
'learning_rate': [0.2, 0.15, .1, 0.05,.05],
'n_nodes': [90, 100, 110],
'batch_size': [80, 90, 100, 110, 120]
}
def create_nn2_hyperas(x_train, x_test, y_train, y_test, params):
seq = Sequential()
seq.add(Dense(output_dim=params['n_nodes'], init="uniform", activation="relu", input_dim=11))
seq.add(Dense(output_dim=1, init="uniform", activation="sigmoid"))
seq.compile(optimizer=Adam(lr=params['learning_rate']), loss="binary_crossentropy", metrics=['accuracy']) # vs categorical_crossentropy
epochs = 200
history = seq.fit(x_train, y_train,
verbose=False,
batch_size=params["batch_size"],
epochs=epochs,
validation_data=[x_test, y_test],
callbacks=[early_stopper(epochs, mode=[0,50])]
)
return history, seq
result = talos.Scan(x_train.values, y_train.values, x_val=x_test.values, y_val=y_test.values, model=create_nn2_hyperas, params=p)
return result
def run_custom_model(my_training_batch_generator, my_val_batch_generator,
input_shape):
params = set_params(input_shape)
verbose = True
round_limit = 2 # NOTE Set this to however many rounds you want to test with
model = custom_model.create_model(my_training_batch_generator,
my_val_batch_generator, verbose)
dummyX, dummyY = my_training_batch_generator.__getitem__(0)
testX, testY = my_val_batch_generator.__getitem__(0)
my_val_batch_generator.on_epoch_end()
tt = talos.Scan(x=dummyX,
y=dummyY,
params=params,
model=model,
x_val=testX,
y_val=testY,
experiment_name='example.csv',
print_params=True,
round_limit=round_limit)
# print(vars(tt), dir(tt))
# print(tt)
t = project_object(tt, 'params', 'saved_models', 'saved_weights', 'data',
'details', 'round_history')
save_object(t, 'example.pickle')
def Optimization():
scan_object = ta.Scan(x=x_train,
y=y_train,
params=parameters,
model=pet_finder_model,
val_split=0,
experiment_name='pet_finder')
# Evaluate
analyze_object = ta.Analyze(scan_object)
scan_data = analyze_object.data
# heatmap correlation
analyze_object.plot_corr('val_accuracy', ['accuracy', 'loss', 'val_loss'])
# a four dimensional bar grid
ast.bargrid(scan_data,
x='lr',
y='val_accuracy',
hue='num_Nodes',
row='loss_function',
col='dropout')
list_of_parameters = analyze_object.table('val_loss',
['accuracy', 'loss', 'val_loss'],
'val_accuracy')
return list_of_parameters
def _train_auto_model(self):
# Param space for Talos model to search through
# Tuples values are a range, Lists are choice options
p = {
'lr': (0.5, 5, 10),
'first_neuron': [24, 48, 96],
'e_size':[128, 300],
'h_size':[32, 64],
'hidden_layers':[2,3],
'activation':[relu],
'batch_size': [64],
'epochs': [2, 5],
'dropout': (0, 0.20, 10),
'optimizer': [Adam],
'seq_len': [self.seq_len],
'vocab_size': [self.vocab_size],
'last_activation': [sigmoid]
}
# Talos scan that will find the best model with the parameters above
h = ta.Scan(self.X_train, self.y_train,
params=p,
model=self.model,
grid_downsample=self.grid_downsample_amount,
val_split=0, # Zerofy val_split in Talos, because we pass in our own val data
x_val=self.X_val,
y_val=self.y_val
)
# Get only the model from the Talos experiment
best_model_index = best_model(h, 'val_acc', False)
self.model = activate_model(h, best_model_index)
def minimal():
x, y = talos.templates.datasets.iris()
p = {
'activation': ['relu', 'elu'],
'optimizer': ['Nadam', 'Adam'],
'losses': ['binary_crossentropy'],
'batch_size': [20, 30, 40],
'epochs': [10, 20]
}
def iris_model(x_train, y_train, x_val, y_val, params):
model = Sequential()
model.add(Dense(32, input_dim=4, activation=params['activation']))
model.add(Dense(3, activation='softmax'))
model.compile(optimizer=params['optimizer'], loss=params['losses'])
out = model.fit(x_train,
y_train,
batch_size=params['batch_size'],
epochs=params['epochs'],
validation_data=[x_val, y_val],
verbose=0)
return out, model
scan_object = talos.Scan(x,
y,
model=iris_model,
params=p,
experiment_name='iris',
fraction_limit=0.1)
return scan_object
def test_latest():
print('\n >>> start Latest Features... \n')
import talos
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
x, y = talos.templates.datasets.iris()
p = {
'activation': ['relu', 'elu'],
'optimizer': ['Nadam', 'Adam'],
'losses': ['logcosh'],
'shapes': ['brick'],
'first_neuron': [16, 32, 64, 128],
'hidden_layers': [0, 1, 2, 3],
'dropout': [.2, .3, .4],
'batch_size': [20, 30, 40, 50],
'epochs': [10]
}
def iris_model(x_train, y_train, x_val, y_val, params):
model = Sequential()
model.add(
Dense(params['first_neuron'],
input_dim=4,
activation=params['activation']))
talos.utils.hidden_layers(model, params, 3)
model.add(Dense(3, activation='softmax'))
model.compile(optimizer=params['optimizer'],
loss=params['losses'],
metrics=['acc'])
out = model.fit(x_train,
y_train,
callbacks=[
talos.utils.ExperimentLogCallback(
'testing_latest', params)
],
batch_size=params['batch_size'],
epochs=params['epochs'],
validation_data=[x_val, y_val],
verbose=0)
return out, model
scan_object = talos.Scan(x,
y,
model=iris_model,
params=p,
experiment_name='testing_latest',
round_limit=5,
reduction_method='gamify',
save_weights=False)
print('finised Latest Features \n')
def optimize(self):
x, y = self.ImageData.images, self.ImageData.labels
def model(x_train, y_train, x_val, y_val, params):
rcnn = RCNN(ImageData=self.ImageData,
loss=self.params['loss'],
opt=self.params['opt'],
lr=lr_normalizer(self.params['lr'],
self.params['opt']),
seed=self.seed,
verbose=0)
return rcnn.train(epochs=self.params['epochs'],
batch_size=self.params['batch_size'],
split_size=self.params['split_size'],
checkpoint_path=None,
early_stopping=False,
verbose=0)
self.scan = talos.Scan(x,
y,
params=self.params,
model=model,
experiment_name='rcnn',
fraction_limit=.001)
return self.scan
def test_predict():
print("\n >>> start Predict()...")
import sys
sys.path.insert(0, '/Users/mikko/Documents/GitHub/talos')
import talos
x, y = talos.templates.datasets.iris()
p = talos.templates.params.iris()
model = talos.templates.models.iris
x = x[:50]
y = y[:50]
scan_object = talos.Scan(x=x,
y=y,
params=p,
model=model,
experiment_name='test_iris',
round_limit=2)
predict = talos.Predict(scan_object)
_preds = predict.predict(x, 'val_acc', False)
_preds = predict.predict_classes(x, 'val_acc', False)
print('finised Predict() \n')
def run(self):
dataset = keras.datasets.mnist
(train_images, train_labels), (test_images,
test_labels) = dataset.load_data()
# standardize the pixel values from 0-255 to 0-1
train_images = util.reshapeSet(train_images)
test_images = util.reshapeSet(test_images)
train_labels = tfku.to_categorical(train_labels, 10)
test_labels = tfku.to_categorical(test_labels, 10)
train_labels = np.asarray(train_labels)
test_labels = np.asarray(test_labels)
if (self.verbose):
print("Train set size: %s" % str(train_images.shape))
print("Train set labels size: %s" % str(train_labels.shape))
print("Test set size: %s" % str(test_images.shape))
print("Test set labels size: %s" % str(test_labels.shape))
print("")
print("First train object <%s>" % train_labels[0])
print("First test object <%s>" % test_labels[0])
print("")
return talos.Scan(train_images,
train_labels,
model=mnist_model,
params=self.params,
x_val=test_images,
y_val=test_labels,
experiment_name=self.name)
def talos_version():
def mnist_model(x_train, y_train, x_val, y_val, params):
model = Sequential()
model.add(Conv2D(32, kernel_size=(3, 3), activation=params['activation'], input_shape=(28, 28, 1)))
model.add(Flatten())
model.add(Dense(128, activation=params['activation']))
model.add(Dropout(params['dropout']))
model.add(Dense(10, activation='softmax'))
model.compile(optimizer=params['optimizer'],
loss=params['losses'],
metrics=['acc', talos.utils.metrics.F1Score()])
out = model.fit_generator(SequenceGenerator(x_train,
y_train,
batch_size=params['batch_size']),
epochs=params['epochs'],
validation_data=[x_val, y_val],
callbacks=[],
workers=4,
verbose=0)
return out, model
scan_object = talos.Scan(x=x_train,
y=y_train,
x_val=x_val,
y_val=y_val,
params=p,
model=mnist_model,
experiment_name='mnist',
save_weights=False)
def get_best_model(x_train, y_train, **kwargs):
np.random.seed(7)
y_pred = kwargs['primal_data']['y_pred']
params = kwargs['params']
params['model_name'] = [kwargs['primal_data']['model_name']]
kwargs.setdefault('dataset_name', 'talos_readings')
kwargs.setdefault('experiment_no', '1')
dataset_name = kwargs['dataset_name']
experiment_no = kwargs['experiment_no']
val_metric = kwargs['val_metric']
metric = kwargs['metric']
for name, value in params.items():
if type(value) != list:
params[name] = [value]
h = ta.Scan(x_train,
y_pred,
params=params,
dataset_name=dataset_name,
experiment_no=experiment_no,
model=talos_model,
grid_downsample=0.5)
report = h.data
best_model = report.sort_values(val_metric, ascending=True).iloc[0]
best_model_id = best_model.name - 1
dnn_model = activate_model(h, best_model_id)
loss = best_model.losses
epochs = int(best_model.epochs)
batch_size = int(best_model.batch_size)
optimizer = best_model.optimizer
dnn_model.compile(optimizer=optimizer, loss=loss, metrics=[metric])
return dnn_model, epochs, batch_size
def hyperparameter_exploration(data, name, num):
"""Hyperparameter exploration with TALOS.
This function explores different hyperparameter combinations with the framework TALOS.
Args:
data -- a dictionary with the training, testing and validation data
name -- name of the experiment
num -- number of the experiment
"""
logger = logging.getLogger('RNN-SA.main.hyperparameter_exploration')
logger.info("Doing hyperparameter exploration...")
start_time = time.time()
talos.Scan(
x=data['train_X'], # prediction features
y=data['train_y'], # prediction outcome variable
params=params.hparams_talos, # the parameter dictionary
model=ml_models.LSTM_model, # the Keras model as a function
dataset_name=name, # used for experiment log
experiment_no=num, # used for experiment log
x_val=data['val_X'], # validation data for x
y_val=data['val_y'], # validation data for y
# grid_downsample=0.1, # a float to indicate fraction for random sampling
print_params=True, # print each permutation hyperparameters
)
end_time = time.time()
logger.info("Finished hyperparameter exploration!")
logger.info("Time elapsed: %f s \n", end_time - start_time)
def explore(self, params):
'''
Explores hyperparameter space and outputs diagnostic plots to file
:param params: A dictionary of hyperparameters to explore
:return: The talos scan history
'''
dir = os.path.join(os.path.dirname(script_dir), hyprOutputDir)
if not os.path.isdir(dir):
os.makedirs(dir)
dir = os.path.join(dir, self.name)
if not os.path.isdir(dir):
os.makedirs(dir)
name = os.path.join(dir, self.name)
x = self.reshape(self.data.getFullX())
y = self.data.getFullY()
h = ta.Scan(x,
y,
params=params,
model=exploration_model,
dataset_name=name,
experiment_no='1',
grid_downsample=.1)
return h
def run_search_cnn(model=search_cnn):
x_train, y_train, x_test, y_test = load_data()
x_train = np.reshape(x_train, (3823, 64, 1))
x_test = np.reshape(x_test, (1797, 64, 1))
p = {
'lr': [0.5, 5],
'kernel': [3, 5],
'batch_size': [32, 16],
'epochs': [100],
'dropout': [0, 0.5],
'activation': ['relu', 'tanh'],
'optimizer': [Adam, Adagrad, RMSprop],
'losses': ['mse', 'categorical_crossentropy'],
'last_activation': ['softmax']
}
# p = {'lr': [0.5, 5],
# 'kernel':[3, 5],
# 'batch_size': [32, 16],
# 'epochs': [100],
# 'dropout': [0, 0.5],
# 'weight_regulizer':[None],
# 'emb_output_dims': [None],
# 'shape':['brick','long_funnel'],
# 'optimizer': [Adam, Adagrad, RMSprop],
# 'losses': ['mse', 'categorical_crossentropy'],
# 'activation':['relu'],
# 'last_activation': ['softmax']}
t = ta.Scan(x=x_train,
y=y_train,
model=model,
experiment_name="cnn",
params=p)
return t
def test_reducers():
print("\n >>> start reducers...")
import talos
x, y = talos.templates.datasets.iris()
p = talos.templates.params.iris()
model = talos.templates.models.iris
x = x[:50]
y = y[:50]
for strategy in [
'trees', 'forrest', 'correlation', 'gamify', 'local_strategy'
]:
talos.Scan(x=x,
y=y,
params=p,
model=model,
experiment_name='test_iris',
round_limit=2,
reduction_method=strategy,
reduction_interval=1)
print('finised reducers \n')
def minimal():
x, y = ta.datasets.iris()
p = {
'activation': ['relu', 'elu'],
'optimizer': ['Nadam', 'Adam'],
'losses': ['logcosh'],
'hidden_layers': [0, 1, 2],
'batch_size': [20, 30, 40],
'epochs': [10, 20]
}
def iris_model(x_train, y_train, x_val, y_val, params):
model = Sequential()
model.add(Dense(32, input_dim=8, activation=params['activation']))
model.add(Dense(1, activation='softmax'))
model.compile(optimizer=params['optimizer'], loss=params['losses'])
out = model.fit(x_train,
y_train,
batch_size=params['batch_size'],
epochs=params['epochs'],
validation_data=[x_val, y_val])
return out, model
scan_object = ta.Scan(x,
y,
model=iris_model,
params=p,
grid_downsample=0.1)
return scan_object
def test_templates():
print("\n >>> start templates ...")
import talos
x, y = talos.templates.datasets.titanic()
x = x[:50]
y = y[:50]
model = talos.templates.models.titanic
p = talos.templates.params.titanic()
talos.Scan(x, y, p, model, 'test', round_limit=2)
x, y = talos.templates.datasets.iris()
x = x[:50]
y = y[:50]
model = talos.templates.models.iris
p = talos.templates.params.iris()
talos.Scan(x, y, p, model, 'test', round_limit=2)
x, y = talos.templates.datasets.cervical_cancer()
x = x[:50]
y = y[:50]
model = talos.templates.models.cervical_cancer
p = talos.templates.params.cervical_cancer()
talos.Scan(x, y, p, model, 'test', round_limit=2)
x, y = talos.templates.datasets.breast_cancer()
x = x[:50]
y = y[:50]
model = talos.templates.models.breast_cancer
p = talos.templates.params.breast_cancer()
talos.Scan(x, y, p, model, 'test', round_limit=2)
x, y = talos.templates.datasets.icu_mortality(50)
x, y = talos.templates.datasets.telco_churn(.3)
x, y, x1, y1 = talos.templates.datasets.mnist()
x, y = talos.templates.datasets.breast_cancer()
x, y = talos.templates.datasets.cervical_cancer()
x, y = talos.templates.datasets.titanic()
talos.templates.pipelines.breast_cancer(random_method='quantum')
talos.templates.pipelines.cervical_cancer(random_method='sobol')
talos.templates.pipelines.iris(random_method='uniform_crypto')
talos.templates.pipelines.titanic(random_method='korobov_matrix')
print("finish templates \n")
def run_optimize(x, y, mode='auto', params=None, grid_downsample=1, epochs=10):
if mode == 'auto':
params = default_params()
params['epochs'] = [epochs]
scan_object = ta.Scan(x,
y,
model=simple_neural_net,
params=params,
grid_downsample=grid_downsample)
else:
scan_object = ta.Scan(x, y, model=simple_neural_net, params=params)
cols = ['val_acc', 'val_loss', 'acc', 'loss']
scan_object.data[cols] = scan_object.data[cols].astype(float).round(3)
return scan_object.data
def search(self, params):
t = ta.Scan(
x=np.zeros((16, 5, 20)), # Add dummy X or talos crashes
y=np.zeros((16, 4)), # Same for y
model=self.trainTalos,
params=params,
experiment_no="1",
)
return t
def iris_pipeline(round_limit=5, random_method='uniform_mersenne'):
'''Performs a Scan with Iris dataset and simple dense net'''
scan_object = ta.Scan(iris()[0],
iris()[1],
iris_params(),
iris_model,
round_limit=round_limit)
return scan_object
def titanic_pipeline(round_limit=2, random_method='uniform_mersenne'):
'''Performs a Scan with Iris dataset and simple dense net'''
scan_object = ta.Scan(titanic()[0][:50],
titanic()[1][:50],
titanic_params(),
titanic_model,
round_limit=round_limit)
return scan_object
def grid_search(self, trainX, trainy, testX, testy, params):
t = ta.Scan(x=trainX,
y=trainy,
x_val=testX,
y_val=testy,
model=self.evaluate_params(),
grid_downsample=0.01,
params=params,
dataset_name='svmONOFF',
experiment_no='1')
def titanic(round_limit=2, random_method='uniform_mersenne'):
'''Performs a Scan with Iris dataset and simple dense net'''
import talos as ta
scan_object = ta.Scan(ta.templates.datasets.titanic()[0][:50],
ta.templates.datasets.titanic()[1][:50],
ta.templates.params.titanic(),
ta.templates.models.titanic,
round_limit=round_limit)
return scan_object
def cervical_cancer(round_limit=2, random_method='uniform_mersenne'):
'''Performs a Scan with Iris dataset and simple dense net'''
import talos as ta
scan_object = ta.Scan(ta.templates.datasets.cervical_cancer()[0],
ta.templates.datasets.cervical_cancer()[1],
ta.templates.params.cervical_cancer(),
ta.templates.models.cervical_cancer,
'test',
round_limit=round_limit)
return scan_object
def _do_scan(data, params, name):
x_train, x_test, y_train, y_test = data
print(params)
talos.Scan(model=mnist_model,
x=x_train,
y=y_train,
x_val=x_test,
y_val=y_test,
params=params,
dataset_name=name,
print_params=True,
clear_tf_session=False)
def run_experiment(time_sec, params, outfile):
X, y = preprocessing('grade_model_training/data/deep_learn_data.csv')
#Set scan paramters
scan_object = ta.Scan(X,
y,
model=import_model,
params=params,
dataset_name="grade_model_training/output/results",
experiment_no=outfile,
grid_downsample=.5)
return scan_object
def run_talosScan(self, fold, last_epoch_value=True, experiment_no="1"):
# and run the experiment
self.boolean = True
t = ta.Scan(x=self.x_train[fold],
y=self.y_train[fold],
model=self.model,
params=self.params,
x_val=self.x_val[fold],
y_val=self.y_val[fold],
last_epoch_value=last_epoch_value,
experiment_no=experiment_no,
search_method='linear')
return t