def train_and_test(self, training_data, test_data):
from pymri.model.ann.theano_script import share_data
training_data, test_data = share_data(training_data, test_data)
# from pymri.model.ann.theano_script import load_data_shared
# training_data, test_data = load_data_shared()
self.net.SGD(
training_data,
self.epochs,
self.mini_batch_size,
self.learning_rate,
test_data,
verbose=self.verbose
)
def train_and_test(self, training_data, test_data):
if 'simple' in self.type:
self.net.SGD(training_data,
self.epochs,
self.mini_batch_size,
self.learning_rate,
test_data=test_data,
verbose=self.verbose)
else:
from pymri.model.ann.theano_script import share_data
training_data, test_data = share_data(training_data, test_data)
self.net.SGD(training_data,
self.epochs,
self.mini_batch_size,
self.learning_rate,
test_data,
verbose=self.verbose)
def train_and_test(self, training_data, test_data):
if 'simple' in self.type:
self.net.SGD(
training_data, self.epochs,
self.mini_batch_size, self.learning_rate,
test_data=test_data,
verbose=self.verbose
)
else:
from pymri.model.ann.theano_script import share_data
training_data, test_data = share_data(training_data, test_data)
self.net.SGD(
training_data,
self.epochs,
self.mini_batch_size,
self.learning_rate,
test_data,
verbose=self.verbose
)
], mini_batch_size)
# ### Dimensionality reduction ################################################
from sklearn.feature_selection import SelectKBest, f_classif
# ### Define the dimension reduction to be used.
# Here we use a classical univariate feature selection based on F-test,
# namely Anova. We set the number of features to be selected to 784
feature_selection = SelectKBest(f_classif, k=300)
# We fit the selector to our tr6ining dataset
feature_selection.fit(X, y)
# Transform training dataset
X = feature_selection.transform(X)
# Transform testing dataset
X_t = feature_selection.transform(X_t)
# ### Train and test classifier ###############################################
# prior chance level
print('Prior chance: %0.2f' % (y_t.sum() / float(y_t.shape[0])))
# prepare data in the format acceptable by theano scripts
from pymri.model.ann.theano_script import share_data
training_data, test_data = share_data((X, y), (X_t, y_t))
net.SGD(training_data, 500, mini_batch_size, 3., test_data, verbose=1)
# from sklearn.metrics import confusion_matrix
# ### Dimensionality reduction ################################################
from sklearn.feature_selection import SelectKBest, f_classif
# ### Define the dimension reduction to be used.
# Here we use a classical univariate feature selection based on F-test,
# namely Anova. We set the number of features to be selected to 784
feature_selection = SelectKBest(f_classif, k=300)
# We fit the selector to our tr6ining dataset
feature_selection.fit(X, y)
# Transform training dataset
X = feature_selection.transform(X)
# Transform testing dataset
X_t = feature_selection.transform(X_t)
# ### Train and test classifier ###############################################
# prior chance level
print('Prior chance: %0.2f' % (y_t.sum()/float(y_t.shape[0])))
# prepare data in the format acceptable by theano scripts
from pymri.model.ann.theano_script import share_data
training_data, test_data = share_data((X, y), (X_t, y_t))
net.SGD(training_data, 500, mini_batch_size, 3., test_data, verbose=1)
# from sklearn.metrics import confusion_matrix