def do_classification():
climate.enable_default_logging()
df = pd.read_csv("glass.csv")
#recode the class variable to go from 0 through 5
df["GT"] = map(recode, df["GT"])
train, valid = train_test_split(df, test_size=0.3)
train_X = train.ix[:, 0:9].as_matrix()
train_Y = train.ix[:, 9].as_matrix()
valid_X = valid.ix[:, 0:9].as_matrix()
valid_Y = valid.ix[:, 9].as_matrix()
train_X = train_X.astype('f')
train_Y = train_Y.astype('i')
valid_X = valid_X.astype('f')
valid_Y = valid_Y.astype('i')
t0 = (train_X, train_Y)
t1 = (valid_X, valid_Y)
exp = theanets.Experiment(theanets.Classifier, layers=(9, 18, 18, 6))
exp.train(t0, t1, algorithm='sgd',\
learning_rate=1e-4, momentum=0.1,\
hidden_l1=0.001, weight_l2=0.001)
cm = confusion_matrix(valid_Y, exp.network.predict(valid_X))
return cm
def __init__(
self,
rank=10,
initializer=np.random.randn,
learning_rate=0.001,
patience=5,
l1_penalty=0.05,
l2_penalty=0.05,
min_improvement=0.005,
max_gradient_norm=5,
optimization_algorithm="adam",
min_value=None,
max_value=None,
verbose=True):
Solver.__init__(
self,
fill_method="zero",
min_value=min_value,
max_value=max_value)
self.rank = rank
self.initializer = initializer
self.learning_rate = learning_rate
self.patience = patience
self.l1_penalty = l1_penalty
self.l2_penalty = l2_penalty
self.max_gradient_norm = max_gradient_norm
self.optimization_algorithm = optimization_algorithm
self.min_improvement = min_improvement
self.verbose = verbose
if self.verbose:
climate.enable_default_logging()
def do_classification():
climate.enable_default_logging()
df = pd.read_csv("glass.csv")
#recode the class variable to go from 0 through 5
df["GT"] = map(recode, df["GT"])
train, valid = train_test_split(df, test_size = 0.3)
train_X = train.ix[:, 0:9].as_matrix()
train_Y = train.ix[:, 9].as_matrix()
valid_X = valid.ix[:, 0:9].as_matrix()
valid_Y = valid.ix[:,9].as_matrix()
train_X = train_X.astype('f')
train_Y = train_Y.astype('i')
valid_X = valid_X.astype('f')
valid_Y = valid_Y.astype('i')
t0 = (train_X, train_Y)
t1 = (valid_X, valid_Y)
exp = theanets.Experiment(theanets.Classifier, layers = (9,18,18,6))
exp.train(t0, t1, algorithm='sgd',\
learning_rate=1e-4, momentum=0.1,\
hidden_l1=0.001, weight_l2=0.001)
cm = confusion_matrix(valid_Y, exp.network.predict(valid_X))
return cm
def do_regression():
climate.enable_default_logging()
train, test = create_datasets()
x_train = train[:, 0:6]
y_train = train[:, 6]
y_train = np.reshape(y_train, (y_train.shape[0], 1))
y_test = test[:, 6]
y_test = np.reshape(y_test, (y_test.shape[0], 1))
exp = theanets.Experiment(theanets.Regressor, layers=(6, 6, 1))
exp.train([x_train, y_train])
# do the testing
x_test = test[:, 0:6]
y_test = test[:, 6]
yp = exp.network.predict(x_test)
xi = [(i + 1) for i in range(x_test.shape[0])]
pb.scatter(xi, y_test, color="red")
pb.scatter(xi, yp, color="blue")
pb.show()
return
def tn_main():
handle_args()
load_type = get_load_type(sys.argv[1])
files = myfiles(PATH_PREFIX, load_type)
# load the matlab files to a dict
inputs = loadMat(files)
# determine the max on all data
maximum = findMax(loadMat(myfiles(PATH_PREFIX, "")))
# extend the x y z coordinates
inputs = extendTo(inputs, maximum)
# split them into train/validation/test datasets (70,20,10)
Tr,V,Tst = create_dataset(inputs)
climate.enable_default_logging()
layer_input = 3 * maximum
layer_hidden = layer_input / 2
layer_output = layer_input / 10
exp = theanets.Experiment(theanets.Classifier, layers=(layer_input, (layer_hidden, 'relu'), (layer_hidden, 'relu') , (layer_output, 'softmax')))
# Train the model
exp.train(Tr, V, learning_rate=1e-4, momentum=0.9, min_improvement=0.01)
np.set_printoptions(threshold='nan')
# Show confusion matrices on the training/validation/test splits.
for label, (X, y) in (('Training:', Tr), ('Validation:', V), ('Test:', Tst)):
predicted = exp.network.predict(X)
result = [(chr(f[0]),chr(f[1])) for f in zip( predicted, y)]
failures = [(b,a) for a,b in result if a != b]
print(label, len(failures),len(y))
print sum([ 1 if (a==b) else 0 for a,b in result ]) / float(len(result))
print "Failures (expected,predicted):{}".format(failures)
print(confusion_matrix(y, predicted))
def __init__(self,
rank=10,
initializer=np.random.randn,
learning_rate=0.001,
patience=5,
l1_penalty=0.05,
l2_penalty=0.05,
min_improvement=0.005,
max_gradient_norm=5,
optimization_algorithm="adam",
min_value=None,
max_value=None,
verbose=True):
Solver.__init__(self,
fill_method="zero",
min_value=min_value,
max_value=max_value)
self.rank = rank
self.initializer = initializer
self.learning_rate = learning_rate
self.patience = patience
self.l1_penalty = l1_penalty
self.l2_penalty = l2_penalty
self.max_gradient_norm = max_gradient_norm
self.optimization_algorithm = optimization_algorithm
self.min_improvement = min_improvement
self.verbose = verbose
if self.verbose:
climate.enable_default_logging()
def do_regression():
climate.enable_default_logging()
train, test = create_datasets()
x_train = train[:, 0:6]
y_train = train[:, 6]
y_train = np.reshape(y_train, (y_train.shape[0], 1))
y_test = test[:, 6]
y_test = np.reshape(y_test, (y_test.shape[0], 1))
exp = theanets.Experiment(theanets.Regressor, layers=(6, 6, 1))
exp.train([x_train, y_train])
#do the testing
x_test = test[:, 0:6]
y_test = test[:, 6]
yp = exp.network.predict(x_test)
xi = [(i + 1) for i in range(x_test.shape[0])]
pb.scatter(xi, y_test, color="red")
pb.scatter(xi, yp, color="blue")
pb.show()
return
def main(input_file, model_path):
batch_size = 128
nb_classes = 62 # A-Z, a-z and 0-9
nb_epoch = 2
# Input image dimensions
img_rows, img_cols = 32, 32
# Path of data files
path = input_file
# Load the preprocessed data and labels
X_train_all = np.load(path + "/trainPreproc_" + str(img_rows) + "_" +
str(img_cols) + ".npy")
Y_train_all = np.load(path + "/labelsPreproc.npy")
X_train, X_val, Y_train, Y_val = \
train_test_split(X_train_all, Y_train_all, test_size=0.25, stratify=np.argmax(Y_train_all, axis=1))
print X_train.shape
labels = convert_(Y_train)
validation = convert_(Y_val)
X_train = X_train.reshape(
(X_train.shape[0], X_train.shape[2] * X_train.shape[3]))
X_val = X_val.reshape((X_val.shape[0], X_val.shape[2] * X_val.shape[3]))
print 'Training...'
class_input = 62
climate.enable_default_logging()
# Build a classifier model with 100 inputs and 10 outputs.
net = theanets.Classifier(layers=[X_train.shape[1], class_input])
X_train = X_train.astype('f')
labels = labels.astype('i')
X_val = X_val.astype('f')
validation = validation.astype('i')
train = X_train, labels
valid = X_val, validation
arg = 'adadelta'
net.train(train,
valid,
algo=arg,
learning_rate=1e-10,
momentum=0.00001,
input_noise=0.3,
hidden_l1=0.1)
print 'saving model paramters to {}'.format(model_path)
with open(model_path, 'wb') as fid:
pickle.dump(net, fid)
print 'Done.'
def main():
training_data, validation_data, test_data, std_scale = load_training_data()
climate.enable_default_logging()
targets = ['esgd','layerwise','rmsprop','nag','rprop','sgd','sample','adadelta']
layers = [(93, dict(size=512, sparsity=0.2, activation='relu'),
dict(size=512, sparsity=0.2, activation='relu'),
dict(size=512, sparsity=0.2, activation='relu'),
9)]
for l in layers:
for t in targets:
exp = theanets.Experiment(
theanets.Classifier,
layers=l,
weighted=True,
output_activation='softmax'
)
exp.train(training_data,
validation_data,
optimize=t,
)
exp.train(training_data,
validation_data,
optimize=t,
)
exp.train(training_data,
validation_data,
optimize=t,
)
#get an prediction of the accuracy from the test_data
test_results = exp.network.predict(test_data[0])
loss = multiclass_log_loss(test_data[1], test_results)
print 'Test multiclass log loss:', loss
out_file = 'results/' + str(loss) + t
exp.save(out_file + '.pkl')
#save the kaggle results
kaggle_test_features = load_test_data(std_scale)
results = exp.network.predict(kaggle_test_features)
save_results(out_file + '.csv', kaggle_test_features, results)
import climate
import ConfigParser
import theano
import theano.tensor as T
import theano.sandbox.rng_mrg
import lasagne
from lasagne.layers import ReshapeLayer, Layer
from lasagne.init import Normal
from lasagne.regularization import regularize_layer_params_weighted, l2, l1
from lasagne.regularization import regularize_layer_params
logging = climate.get_logger('trainer')
climate.enable_default_logging()
def load_model(filename):
f = file(filename, 'rb')
params = cPickle.load(f)
f.close()
return params
def save_model(filename, model):
params = lasagne.layers.get_all_param_values(model)
f = file(filename, 'wb')
cPickle.dump(params, f, protocol=cPickle.HIGHEST_PROTOCOL)
f.close()
return None
def experiment(
X_learn, y_learn, X_test, y_test,
algo='rmsprop',
learning_rate=0.0001,
momentum=0,
neurons=10**2,
patience=100*1000,
# Sparse hidden activations
# have shown much promise in computational neural networks.
hidden_l1 = 0.01,
weight_l2 = 0.0001,
):
import climate
climate.enable_default_logging()
TRAIN_PART = 0.7
data_threshold = int(TRAIN_PART * len(X_learn))
print 'Divide sets...'
np.random.seed(2016)
np.random.shuffle(X_learn)
np.random.seed(2016)
np.random.shuffle(y_learn)
X_train, y_train = X_learn[:data_threshold], y_learn[:data_threshold]
X_valid, y_valid = X_learn[data_threshold:], y_learn[data_threshold:]
datasets = {
'training': (X_train, y_train),
'validation': (X_valid, y_valid),
}
layers=(X_learn.shape[1], (neurons, 'relu'), 2)
import theanets
exp = theanets.Experiment(
theanets.Classifier,
layers=layers,
)
train_acc_history = []
valid_acc_history = []
valid_0_acc_history = []
valid_1_acc_history = []
init_learning_plot()
from sklearn.metrics import accuracy_score
print 'Start learning...'
iteration = 0
for tm, vm in exp.itertrain(
datasets['training'],
datasets['validation'],
algo=algo,
learning_rate=learning_rate,
momentum=momentum,
hidden_l1=hidden_l1,
weight_l2=weight_l2,
patience=patience,
):
iteration += 1
# Validate every class separately
y_pred_0 = exp.network.classify(X_valid[y_valid == 0])
y_pred_1 = exp.network.classify(X_valid[y_valid == 1])
valid_0_acc_history.append(
accuracy_score(y_valid[y_valid == 0], y_pred_0) * 100)
valid_1_acc_history.append(
accuracy_score(y_valid[y_valid == 1], y_pred_1) * 100)
train_acc_history.append(tm['acc'] * 100)
valid_acc_history.append(vm['acc'] * 100)
# FIXME: First validation ACC is 1.0
update_learning_plot(
train_acc_history[10:],
valid_acc_history[10:],
valid_0_acc_history[10:],
valid_1_acc_history[10:])
if iteration == 490:
save_learning_plot()
save_learning_plot('current_learning_end.png')
from sklearn.metrics import classification_report, confusion_matrix
y_pred = exp.network.classify(X_test)
print 'classification_report:\n', classification_report(y_test, y_pred)
print 'confusion_matrix:\n', confusion_matrix(y_test, y_pred)
THEANO_FLAGS = 'device=cpu,device=gpu0'
THEANO_FLAGS = 'floatX=float32,device=gpu0,lib.cnmem=1'
import theanets
import climate
import numpy as np
import pandas as pd
import math
from sklearn import preprocessing
from datetime import datetime
climate.enable_default_logging() # to print downhill's iteration result
path = 'C:/Users/Administrator/Desktop/Data'
SNR2 = [0, 5, 10]
MFCC = 22
X_train = [None] * MFCC
X_train = np.array(np.float64(X_train))
# print X_train.shape
start0 = datetime.now()
print('loading MAG_y_train...')
for n in range(1, 6):
for level in SNR2:
X_train_0 = np.genfromtxt(path + '/MFCCs_train' + '_SNR_' +
str(level) + '_noise_' + str(n) + '.csv',
delimiter=',')
X_train = np.vstack((X_train, X_train_0))
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import division, print_function, absolute_import
from sklearn.preprocessing.data import StandardScaler
from rep.test.test_estimators import check_classifier, check_regression, check_params, \
check_classification_reproducibility
from rep.test.test_estimators import generate_classification_data
from sklearn.metrics import roc_auc_score
from sklearn.ensemble import BaggingClassifier
from rep.estimators.sklearn import SklearnClassifier
from rep.estimators.theanets import TheanetsClassifier, TheanetsRegressor
import climate
from tests import known_failure, retry_if_fails
climate.enable_default_logging(default_level='ERROR')
__author__ = 'Lisa Ignatyeva, Tatiana Likhomanenko, Alex Rogozhnikov'
classifier_params = {
'has_staged_pp': False,
'has_importances': False,
'supports_weight': True,
}
regressor_params = {
'has_staged_predictions': False,
'has_importances': False,
'supports_weight': True,
}
def experiment(
X_learn,
y_learn,
X_test,
y_test,
neurons,
theanets_kwargs,
column_names=None,
name='my_exp',
max_iters=float('inf'),
gather_metrics=[
'accuracy_score',
'matthews_corrcoef',
'roc_auc_score',
'f1_score',
],
plot_every=50,
regression=False,
):
import climate
climate.enable_default_logging()
TRAIN_PART = 0.7
data_threshold = int(TRAIN_PART * len(X_learn))
print 'Divide sets...'
np.random.seed(2016)
np.random.shuffle(X_learn)
np.random.seed(2016)
np.random.shuffle(y_learn)
X_train, y_train = X_learn[:data_threshold], y_learn[:data_threshold]
X_valid, y_valid = X_learn[data_threshold:], y_learn[data_threshold:]
datasets = {
'training': (X_train, y_train),
'validation': (X_valid, y_valid),
'0 validation': (X_valid[y_valid == 0], y_valid[y_valid == 0]),
'1 validation': (X_valid[y_valid == 1], y_valid[y_valid == 1]),
'0 test': (X_test[y_test == 0], y_test[y_test == 0]),
'1 test': (X_test[y_test == 1], y_test[y_test == 1]),
'test': (X_test, y_test)
}
for key in datasets.keys():
datasets[key] = (datasets[key][0], datasets[key][1])
layers = (X_learn.shape[1], ) + neurons + (2, )
if regression:
layers = (X_learn.shape[1], ) + neurons + (1, )
import theanets
exp = theanets.Experiment(
theanets.Classifier,
layers=layers,
)
if regression:
exp = theanets.Experiment(theanets.Regressor, layers=layers)
network = exp.network
print 'Start learning...'
iteration = 0
# Metrics that we want to measure while observing results
metrics = []
# ACC
if 'accuracy_score' in gather_metrics:
metrics.append({
'name': 'accuracy_score',
'function': sklearn.metrics.accuracy_score
})
# MCC
if 'matthews_corrcoef' in gather_metrics:
metrics.append({
'name': 'matthews_corrcoef',
'function': sklearn.metrics.matthews_corrcoef
})
# AUC
if 'roc_auc_score' in gather_metrics:
metrics.append({
'name': 'roc_auc_score',
'function': sklearn.metrics.roc_auc_score
})
# F1
if 'f1_score' in gather_metrics:
metrics.append({
'name': 'f1_score',
'function': sklearn.metrics.f1_score
})
# Regression
# MSE
if 'mse' in gather_metrics:
metrics.append({
'name': 'mse',
'function': sklearn.metrics.mean_squared_error
})
if 'r2_score' in gather_metrics:
metrics.append({
'name': 'r2_score',
'function': sklearn.metrics.r2_score
})
if column_names is not None:
for w_i in xrange(X_learn.shape[1]):
metrics.append({
'name': 'hid1_w_' + column_names[w_i],
'inspect': ('hid1', 'w', w_i)
})
# Sets that we want to test
plot_sets = ['training', 'validation', '0 test', '1 test', 'test']
init_learning_plot()
# Init histories
plot_history = dict()
for metric in metrics:
plot_history[metric['name']] = dict()
for plot_set in plot_sets:
if metric['name'] == 'roc_auc_score' \
and plot_set[0] == '0':
continue
if metric['name'] == 'roc_auc_score' \
and plot_set[0] == '1':
continue
if metric['name'] == 'f1_score' \
and plot_set[0] == '0':
continue
if metric['name'] == 'f1_score' \
and plot_set[0] == '1':
continue
if metric['name'] == 'matthews_corrcoef' \
and plot_set[0] == '0':
continue
if metric['name'] == 'matthews_corrcoef' \
and plot_set[0] == '1':
continue
plot_history[metric['name']][plot_set] = []
iters = []
# Automatic metrics from theanets
#plot_history['error'] = dict()
#plot_history['error']['validation'] = []
#plot_history['loss'] = dict()
#plot_history['loss']['validation'] = []
# XXX
#plot_history['accuracy_score']['training'] = []
print 'Collecting following metrics:'
print gather_metrics
for tm, vm in exp.itertrain(datasets['training'], datasets['validation'],
**theanets_kwargs):
iteration += 1
if iteration > max_iters:
break
if iteration % plot_every:
continue
for metric in metrics:
if not metric['name'] in gather_metrics:
continue
for plot_set in plot_sets:
X_set, y_set = datasets[plot_set]
if metric['name'] == 'roc_auc_score' \
and plot_set[0] == '0':
continue
if metric['name'] == 'roc_auc_score' \
and plot_set[0] == '1':
continue
if metric['name'] == 'f1_score' \
and plot_set[0] == '0':
continue
if metric['name'] == 'f1_score' \
and plot_set[0] == '1':
continue
if metric['name'] == 'matthews_corrcoef' \
and plot_set[0] == '0':
continue
if metric['name'] == 'matthews_corrcoef' \
and plot_set[0] == '1':
continue
if metric['name'] == 'accuracy_score' \
and plot_set == 'validation':
# Get it directly from theanets
plot_history[metric['name']][plot_set].append(vm['acc'] *
100)
continue
if metric['name'] == 'accuracy_score' \
and plot_set == 'training':
# Get it directly from theanets
plot_history[metric['name']]['training'].append(tm['acc'] *
100)
continue
if not regression:
y_pred = exp.network.classify(X_set)
else:
y_pred = exp.network.predict(X_set)
if 'function' in metric:
scaling = 100.
mini = 0.
if metric['name'] == 'mse' \
or metric['name'] == 'r2_score':
scaling = 1.
mini = float('-inf')
plot_history[metric['name']][plot_set].append(
max(mini, metric['function'](y_set, y_pred) * scaling))
if 'inspect' in metric:
layer, param_name, param_x = metric['inspect']
param = network.find(layer, param_name)
values = param.get_value()
mean_value = np.mean(values[param_x]) * 100.
plot_history[metric['name']][plot_set].append(mean_value)
iters.append(iteration)
print 'iteration', iteration, {
m + '_' + s: plot_history[m][s][-1:]
for m in plot_history for s in plot_history[m]
}
update_learning_plot(iters, plot_history)
if iteration in [500, 1000, 2000, 5000]:
save_learning_plot('current_learning_' + str(iteration) + '_' +
name + '.png')
save_learning_plot('current_learning_end.png')
from sklearn.metrics import classification_report, confusion_matrix
if not regression:
y_pred = exp.network.classify(X_test)
else:
y_pred = exp.network.predict(X_test)
if not regression:
print 'classification_report:\n', \
classification_report(y_test, y_pred)
print 'confusion_matrix:\n', \
confusion_matrix(y_test, y_pred)
for metric in metrics:
plot_history[metric['name']]['test_max'] = max(
plot_history[metric['name']]['test'])
return plot_history
def run(path_project,
path_analysis,
cfg_project,
cfg_ann,
sgls_all,
plots=False,
debug=False):
'''
Compile subglide data, tune network architecture and test dataset size
Args
----
cfg_project: OrderedDict
Dictionary of configuration parameters for the current project
cfg_ann: OrderedDict
Dictionary of configuration parameters for the ANN
debug: bool
Swith for running single network configuration
plots: bool
Switch for generating diagnostic plots after each network training
Returns
-------
cfg: dict
Dictionary of network configuration parameters used
data: tuple
Tuple collecting training, validation, and test sets. Also includes bin
deliniation values
results: tuple
Tuple collecting results dataframes and confusion matrices
Note
----
The validation set is split into `validation` and `test` sets, the
first used for initial comparisons of various net configuration
accuracies and the second for a clean test set to get an true accuracy,
as reusing the `validation` set can cause the routine to overfit to the
validation set.
'''
from collections import OrderedDict
import climate
import numpy
import os
import pandas
import theano
import yamlord
from . import utils_ann
from .utils_ann import ppickle
from ..config import paths, fnames
# Environment settings - logging, Theano, load configuration, set paths
#---------------------------------------------------------------------------
climate.enable_default_logging()
theano.config.compute_test_value = 'ignore'
# Configuration settings
if debug is True:
for key in cfg_ann['net_tuning'].keys():
cfg_ann['net_tuning'][key] = [
cfg_ann['net_tuning'][key][0],
]
# Drop fields missing values
sgls_nonan = sgls_all.dropna()
print('\nSplit and normalize input/output data')
features = cfg_ann['net_all']['features']
target = cfg_ann['net_all']['target']
n_targets = cfg_ann['net_all']['n_targets']
valid_frac = cfg_ann['net_all']['valid_frac']
# Normalize input (features) and output (target)
nsgls, bins = _normalize_data(sgls_nonan, features, target, n_targets)
# Get indices of train, validation and test datasets
ind_train, ind_valid, ind_test = _split_indices(nsgls, valid_frac)
# Split dataframes into train, validation and test (features, targets) tuples
train, valid, test = _create_datasets(nsgls, ind_train, ind_valid,
ind_test, features, target)
print('train', len(train[0]), len(train[1]))
print('valid', len(valid[0]), len(valid[1]))
print('test', len(test[0]), len(test[1]))
# Save information on input data to config
cfg_ann['net_all']['targets'] = [float(b) for b in bins]
# Tuning - find optimal network architecture
#---------------------------------------------------------------------------
print('\nTune netork configuration')
# Get all dict of all configuration permutations of params in `tune_params`
configs = _get_configs(cfg_ann['net_tuning'])
# Cycle through configurations storing configuration, net in `results_tune`
n_features = len(cfg_ann['net_all']['features'])
n_targets = cfg_ann['net_all']['n_targets']
print('\nNumber of features: {}'.format(n_features))
print('Number of targets: {}\n'.format(n_targets))
results_tune, tune_accuracy, cms_tune = _tune_net(
train,
valid,
test,
bins,
configs,
n_features,
n_targets,
plots,
)
# Get neural net configuration with best accuracy
best_config = get_best(results_tune, 'config')
# Test effect of dataset size
#---------------------------------------------------------------------------
print('\nRun percentage of datasize tests')
# Get randomly sorted and subsetted datasets to test effect of dataset_size
# i.e. - a dataset with the first `subset_fraction` of samples.
results_dataset, data_accuracy, cms_data = _test_dataset_size(
best_config, train, valid, test, bins, n_features, n_targets, plots,
debug)
print('\nTest data accuracy (Configuration tuning): {}'.format(
tune_accuracy))
print(
'Test data accuracy (Datasize test): {}'.format(data_accuracy))
# Save results and configuration to output directory
#---------------------------------------------------------------------------
# Create output directory if it does not exist
path_output = os.path.join(path_project, paths['ann'], path_analysis)
os.makedirs(path_output, exist_ok=True)
# Save updated `cfg_ann` to output directory
file_cfg_ann = os.path.join(path_output, fnames['cfg']['ann'])
yamlord.write_yaml(cfg_ann, os.path.join(path_output, file_cfg_ann))
# Compiled SGLs before NaN drop and normalization
utils_ann.ppickle(sgls_all, os.path.join(path_output,
fnames['ann']['sgls']))
# Compiled SGLs after NaN drop and normalization
utils_ann.ppickle(nsgls,
os.path.join(path_output, fnames['ann']['sgls_norm']))
# Save output data to analysis output directory
tune_fname = fnames['ann']['tune']
datasize_fname = fnames['ann']['dataset']
ppickle(results_tune, os.path.join(path_output, tune_fname))
ppickle(results_dataset, os.path.join(path_output, datasize_fname))
# Save train, validation, test datasets
ppickle(train, os.path.join(path_output, fnames['ann']['train']))
ppickle(valid, os.path.join(path_output, fnames['ann']['valid']))
ppickle(test, os.path.join(path_output, fnames['ann']['test']))
ppickle(cms_tune, os.path.join(path_output, fnames['ann']['cms_tune']))
ppickle(cms_data, os.path.join(path_output, fnames['ann']['cms_data']))
return cfg_ann, (train, valid, test), (results_tune, results_dataset,
cms_tune, cms_data)
#!/usr/bin/env python
import climate
import matplotlib.pyplot as plt
import numpy as np
import numpy.random as rng
import theanets
climate.enable_default_logging()
S = np.linspace(0, 4 * np.pi, 256)
def wave(i=0):
return (0.4 * np.sin(S) + 0.3 * np.sin(i * S / 2))[:, None, None]
def waves(n=64):
return np.concatenate([wave(rng.randint(15, 30)) for _ in range(n)], axis=1).astype('f')
# set up a network and train it using some sinusoidal data.
e = theanets.Experiment(
theanets.recurrent.Regressor,
layers=(2, 10, 1),
train_batches=16)
def sum_waves():
x = waves()
y = waves()
return [np.concatenate([x, y], axis=2), x + y]
e.run(sum_waves, sum_waves)
# limitations under the License.
from __future__ import division, print_function, absolute_import
from sklearn.preprocessing.data import StandardScaler
from rep.test.test_estimators import check_classifier, check_regression, check_params, \
check_classification_reproducibility
from rep.test.test_estimators import generate_classification_data
from sklearn.metrics import roc_auc_score
from sklearn.ensemble import BaggingClassifier
from rep.estimators.sklearn import SklearnClassifier
from rep.estimators.theanets import TheanetsClassifier, TheanetsRegressor
import climate
from tests import known_failure, retry_if_fails
climate.enable_default_logging(default_level='ERROR')
__author__ = 'Lisa Ignatyeva, Tatiana Likhomanenko, Alex Rogozhnikov'
classifier_params = {
'has_staged_pp': False,
'has_importances': False,
'supports_weight': True,
}
regressor_params = {
'has_staged_predictions': False,
'has_importances': False,
'supports_weight': True,
}
