def get_dataset(dir, win_size=5, is_negful=True):
input_data = InputData(dir, window_size=win_size)
cv_features, cv_labels, extra_negatives, extra_negative_labels = input_data.getData(
)
if is_negful:
return cross_validation.CrossValidation(
cv_features,
cv_labels,
extra_negatives=extra_negatives,
extra_labels=extra_negative_labels)
else:
return cross_validation.CrossValidation(cv_features, cv_labels)
def test_writeFolds(self):
cs = cross_validation.CrossValidation(self.raddress, self.waddress, 3)
cs.readFolder()
cs.writeFolder()
testSet = [
'/staf/amir/robotica/Brain/data/hog_test/cswrite/1.jpg',
'/staf/amir/robotica/Brain/data/hog_test/cswrite/1320757889.jpg'
]
trainingSet = [
'/staf/amir/robotica/Brain/data/hog_test/cswrite/4.jpg',
'/staf/amir/robotica/Brain/data/hog_test/cswrite/5.jpg'
]
for i in range(3):
cs.writeFolds(i, testSet, trainingSet)
testList = os.listdir(
'/staf/amir/robotica/Brain/data/hog_test/cswrite/' + str(i) +
"/testset")
trainingList = os.listdir(
'/staf/amir/robotica/Brain/data/hog_test/cswrite/' + str(i) +
"/trainingset")
self.assertNotEqual(testList, [], "Test directory is empty")
self.assertNotEqual(trainingList, [],
"Training directory is empty")
for tests in testList:
for trains in trainingList:
self.assertNotEqual(
tests, trains,
"The file in the test set is equal to the file in train set"
)
def get_test_dateset(dir, site='Y', win_size=5, is_context=False):
input_data = TestData(dir,
site,
window_size=win_size,
is_context=is_context)
ids, seqs, feature, label = input_data.getData()
return ids, seqs, cross_validation.CrossValidation(
feature, label), np.concatenate(label, axis=0)
def test_writeFolder(self):
cs = cross_validation.CrossValidation(self.raddress, self.waddress,
self.k_fold)
cs.readFolder()
data = cs.writeFolder()
self.assertEqual(
9, data,
"The number of files are not equal to the actual number of files in folders"
)
def test_readFolder(self):
cs = cross_validation.CrossValidation(self.raddress, self.waddress,
self.k_fold)
data = cs.readFolder()
self.assertEqual(
len(data), 9,
"The number of classes is not equal to the number of folders")
self.assertEqual(
os.path.isfile(os.path.join(self.raddress, 'data-label.dat')),
True, "The data-label file is not written.")
def test_dataShuffler(self):
testData = [
'/staf/amir/test/1.png', '/staf/amir/test/2.png',
'/staf/amir/test/3.png', '/staf/amir/test/2.png',
'/staf/amir/test/3.png', '/staf/amir/test/2.png',
'/staf/amir/test/3.png'
]
length = len(testData)
cs = cross_validation.CrossValidation(self.raddress, self.waddress,
self.k_fold)
testShuffle = cs.dataShuffler(testData)
self.assertEqual(length, len(testShuffle),
"The lenght of the input and output differs")
def test_setFolds(self):
testSliced = [
['/staf/amir/robotica/Brain/data/hog_test/cswrite/1.jpg'],
['/staf/amir/robotica/Brain/data/hog_test/cswrite/4.jpg'],
['/staf/amir/robotica/Brain/data/hog_test/cswrite/5.jpg']
]
cs = cross_validation.CrossValidation(self.raddress, self.waddress, 3)
cs.readFolder()
cs.writeFolder()
set = cs.setFolds(testSliced)
for i in range(3):
testList = os.listdir(
'/staf/amir/robotica/Brain/data/hog_test/cswrite/' + str(i) +
"/testset")
trainingList = os.listdir(
'/staf/amir/robotica/Brain/data/hog_test/cswrite/' + str(i) +
"/trainingset")
for tests in testList:
for trains in trainingList:
self.assertNotEqual(
tests, trains,
"The file in the test set is equal to the file in train set"
)
def test_dataOrganizer(self):
cs = cross_validation.CrossValidation(self.raddress, self.waddress,
self.k_fold)
cs.readFolder()
cs.writeFolder()
cs.dataOrganizer()
else:
x_source_tf = x_source
transformed = False
# Scale data
scaler = StandardScaler()
scale_data = True
if scale_data:
print 'scale data'
scaler.fit(x_source_tf)
x_source_tf = scaler.transform(x_source_tf)
# Cross validation
data_cv = np.hstack((ids, y_source, x_source_tf))
cross_validation = cv.CrossValidation(data_cv, 3)
# switch on/off if cross validation or test data prediction
cross_validate = True
if cross_validate:
print 'Doing Cross Validation'
param_manager = ParameterManager()
myrange = [5 * x for x in range(1, 5)]
# Define parameters here: (parameter_name, [parameter_values])
parameter_settings = [('alpha', myrange),
('layer_size', [(100, 100, 100, 100, 100, 100, 100,
100), (50)]),
if __name__ == "__main__":
if len(sys.argv) < NUM_ARGS + 1:
print _usage()
sys.exit(1)
data_file = sys.argv[1]
handle = open(data_file, 'r')
handle.readline()
csv_file = csv.reader(handle)
data = []
for line in csv_file:
l = tuple(map(lambda x: float(x), line[0:2]))
data.append(l)
weight_kernel = 1
weight_uniform = 1
cv = cross_validation.CrossValidation(NUM_FOLDS, data, True)
for i in xrange(NUM_FOLDS):
N = cv.num_training_examples(i)
data = cv.training_examples
for t in cv.training_examples(i):
k_log_likelihood = kernel_log_likelihood(t, i, data, N)
u_log_likelihood = uniform_log_likelihood(t)
if k_log_likelihood > u_log_likelihood:
weight_kernel += 1
else:
weight_uniform += 1
print weight_kernel, weight_uniform, k_log_likelihood, u_log_likelihood
y_source = source_tf[:, 0].reshape(source_tf.shape[0], 1)
else:
x_source_tf = feature_transform(feature_vec, x_source)
else:
x_source_tf = x_source
else:
x_source_tf = x_source
print ids.shape
print y_source.shape
print x_source_tf.shape
data_cv = np.hstack((ids, y_source, x_source_tf))
cross_validation = cv.CrossValidation(data_cv, int(20))
# lin_reg = lr.LinearRegression()
# cross validate over lamda in ridge regression
cross_validate = False
if cross_validate:
print 'Doing Cross Validation'
results = []
scale = 0.02
scale_h = 1.
for i in range(10, 20, 1):
print i
for j in [1]:
ga.evolve(freq_stats=1)
print ga.bestIndividual()
best_chromosome = ga.bestIndividual()
return best_chromosome
if __name__ == "__main__":
arguments = docopt(__doc__)
method = arguments['<method>']
bindres_file = arguments['<binding_residue_file>']
pssms_file = arguments['<pssms_file>']
log_file = arguments['<log_file>']
output_file = arguments['<output_file>']
crossValidation = cross_validation.CrossValidation(bindres_file,
pssms_file, log_file,
method)
best_chromosome = run_ga(crossValidation)
with open(output_file, "w") as fp:
if crossValidation.method == "neuralNetwork":
fp.write(
"#method\tnode_num\tlearning_rate\twindow_size\tdecision_value\n"
)
elif crossValidation.method == "randomForest":
fp.write(
"#method\tn_estimators\tmax_features\twindow_size\tdecision_value\n"
)
elif crossValidation.method == "SVM":
fp.write("#method\tcost\tgamma\twindow_size\tdecision_value\n")
gene1, gene2, gene3 = crossValidation.decode_chromosome(
best_chromosome)