def __init__(self):
self.hog_num_file = "ann/input/hog-nums.ann"
self.hog_sym_file = "ann/input/hog-syms.ann"
self.num_ann = lf.neural_net()
self.sym_ann = lf.neural_net()
self.num_ann.create_from_file(self.hog_num_file)
self.sym_ann.create_from_file(self.hog_sym_file)
print 'loading ANN...'
def __init__(self):
self.hog_num_file="ann/input/hog-nums.ann"
self.hog_sym_file = "ann/input/hog-syms.ann"
self.num_ann = lf.neural_net()
self.sym_ann = lf.neural_net()
self.num_ann.create_from_file(self.hog_num_file)
self.sym_ann.create_from_file(self.hog_sym_file)
print 'loading ANN...'
def test_network(file_name):
# Grab testing file from the user
test_file = input('\nWhat file do you want to test on?: ')
# Now read from the file
file = open(test_file, 'r')
# Loop through the file line by line and grab the input and outputs
inputs = []
outputs = []
num = 0
for line in file:
array = line.split(',')
nums = [int(i) for i in array]
if num % 2 == 0:
inputs.append(nums)
else:
outputs.append(nums)
num += 1
ann = nn.neural_net()
ann.create_from_file(file_name)
# Now test the network
correct = 0
for i, inpt in enumerate(inputs):
answer = ann.run(inpt)
answer = math.ceil(outputs[i][0] - answer[0])
if answer == outputs[i][0]:
correct += 1
print('\nAccuracy is ', (correct / len(outputs)) * 100, '%\n')
def select_train_set(weibo_items):
start_time = timeit.default_timer()
"""
Randomly Select Train Set
"""
# 75% total
train_list = random.sample(weibo_items, len(weibo_items) * 3 / 4)
# 25% for test
test_list = []
for item in weibo_items:
if item not in train_list:
test_list.append(item)
train_set_lines = ''
# First line: #pairs #inputs #output
train_set_lines += str(len(train_list)) + ' 3 1\n'
# [1]emotion grate [2]time [3]forward
# [4]suicide or not
for item in weibo_items:
train_set_lines += str(item[1]) + ' ' + str(item[2]) + ' ' + str(item[3]) + ' \n'
train_set_lines += str(item[4]) + '\n'
# Create neural network folder
if not(os.path.exists('neural-network')):
os.mkdir('neural-network')
# /Users/LemonC/Code/Python/WeiboPrediction/neural-network
os.chdir(os.path.abspath(os.curdir + '/neural-network'))
# Write input file for neural network
with open('NN-train-set.txt', 'w') as fout_for_NN:
fout_for_NN.write(train_set_lines)
print('Prepare train set for Neural Network successfully')
"""
Neural Network Training
"""
connection_rate = 1 # full connect
learning_rate = 0.7
number_input = 3
number_hidden = 5
number_output = 1
desired_error = 0.001
max_iterations = 10000
iterations_between_reports = 1000
ann = libfann.neural_net()
ann.create_sparse_array(connection_rate, (number_input, number_hidden, number_output))
ann.set_learning_rate(learning_rate)
ann.set_activation_function_hidden(libfann.SIGMOID_SYMMETRIC) # (-1, 1)
ann.set_activation_function_output(libfann.SIGMOID) # (0,1)
ann.train_on_file('NN-train-set.txt', max_iterations, iterations_between_reports, desired_error)
ann.save('trained.net')
ann.destroy()
stop_time = timeit.default_timer()
print('Total training time: ' + str('%.2f' % (stop_time - start_time) + ' seconds.\n'))
# Back to root folder
os.chdir('..')
return test_prediction(test_list)
def configure_net(self):
self.net = fann.neural_net()
self.net.create_standard_array([len(self.ids), 10, 1])
self.net.set_activation_function_hidden(fann.SIGMOID_SYMMETRIC_STEPWISE)
self.net.set_activation_function_output(fann.SIGMOID_SYMMETRIC_STEPWISE)
self.net.set_train_stop_function(fann.STOPFUNC_BIT)
self.net.set_bit_fail_limit(0.1)
def test_anns(results):
data = read_metrics(results)
split = int(len(data) * 0.8)
num_input = len(data[0]) - 1
random.shuffle(data)
test_data = data[split:]
# Read ANNs
ann_dir = '/home/tomas/Dropbox/Git/ga_sandbox/projects/denoising/neural/trained_anns'
trained_anns = []
for filename in os.listdir(ann_dir):
if filename.endswith('.net'):
ann_path = os.path.join(ann_dir, filename)
ann = libfann.neural_net()
ann.create_from_file(ann_path)
trained_anns.append(ann)
points = []
for row in test_data:
actual_output = row[num_input]
ann_mean_output = np.mean([
ann.run(row[:num_input])
for ann in trained_anns
])
points.append([ann_mean_output, actual_output])
print "actual: " + str(actual_output) + ", predicted: " + str(ann_mean_output)
points = sorted(points, key=lambda p: p[1])
fig = plt.figure(figsize=(6, 4))
plt.plot([p[0] for p in points])
plt.plot([p[1] for p in points])
plt.ylim([0, 1.2])
plt.show()
def fann_init_net(self,
hidden_func=libfann.SIGMOID_SYMMETRIC_STEPWISE,
output_func=libfann.SIGMOID_STEPWISE):
net = libfann.neural_net()
net.set_activation_function_hidden(hidden_func)
net.set_activation_function_output(output_func)
return net
def from_file(cls, name, prefix):
prefix += '.intent'
self = cls(name)
self.net = fann.neural_net()
self.net.create_from_file(str(prefix + '.net')) # Must have str()
self.ids.load(prefix)
return self
def recreate_images(result_dir, noisy_image_dir):
# Read noisy images first
test_images = {}
for image_name in os.listdir(noisy_image_dir):
if image_name.endswith('.png'):
image_path = os.path.join(noisy_image_dir, image_name)
image = util.img_as_float(io.imread(image_path))
image_name_noext = os.path.splitext(image_name)[0]
test_images[image_name_noext] = image
# Enumerate results - image directories
for image_name in sorted(os.listdir(result_dir)):
image_dir = os.path.join(result_dir, image_name)
if os.path.isdir(image_dir):
print image_name
for result_file in sorted(os.listdir(image_dir)):
if result_file.endswith('.net'):
# Instantiate trained ANN from .net file
net_path = os.path.join(image_dir, result_file)
ann = libfann.neural_net()
ann.create_from_file(net_path)
# Filter the same image which it was trained with
filtered_image = filter_fann(
test_images[image_name], ann)
param_set_name = os.path.splitext(result_file)[0]
io.imsave(
os.path.join(image_dir, param_set_name + '.png'),
filtered_image)
def train(self, train_data):
self.set_train_data(train_data)
hidden_layers = [self.hidden_neurons] * self.hidden_layers
layers = [self.train_data.num_input]
layers.extend(hidden_layers)
layers.append(self.train_data.num_output)
sys.stderr.write("Network layout:\n")
sys.stderr.write("* Neuron layers: %s\n" % layers)
sys.stderr.write("* Connection rate: %s\n" % self.connection_rate)
if self.training_algorithm not in ('TRAIN_RPROP',):
sys.stderr.write("* Learning rate: %s\n" % self.learning_rate)
sys.stderr.write("* Activation function for the hidden layers: %s\n" % self.activation_function_hidden)
sys.stderr.write("* Activation function for the output layer: %s\n" % self.activation_function_output)
sys.stderr.write("* Training algorithm: %s\n" % self.training_algorithm)
self.ann = libfann.neural_net()
self.ann.create_sparse_array(self.connection_rate, layers)
self.ann.set_learning_rate(self.learning_rate)
self.ann.set_activation_function_hidden(getattr(libfann, self.activation_function_hidden))
self.ann.set_activation_function_output(getattr(libfann, self.activation_function_output))
self.ann.set_training_algorithm(getattr(libfann, self.training_algorithm))
fann_train_data = libfann.training_data()
fann_train_data.set_train_data(self.train_data.get_input(), self.train_data.get_output())
self.ann.train_on_data(fann_train_data, self.epochs, self.iterations_between_reports, self.desired_error)
return self.ann
def main():
args = parse_args()
print "loading data"
x_pca_test = np.load("data/x-test-pca%s_%s.npy" % (args.pca, args.type))
x_test = np.load("data/x-test_%s.npy" % args.type)
y_test = np.load("data/y-test_%s.npy" % args.type)
print "x_pca_test Shape", x_pca_test.shape
print "x_test Shape", x_test.shape
print "y_test Shape", y_test.shape
print "Loading fann"
ann = libfann.neural_net()
ann.create_from_file("nnets/%s" % args.fann_file)
predicted = []
for test in x_pca_test:
predicted.append(np.array(ann.run(test)))
print "Test Error:", get_error(predicted, y_test)
print "Saving predictions"
np.save("data/p-test_%s" % args.type, np.array(predicted))
print "Smoothing predictions"
predicted_smooth = smooth(args, predicted)
print "Smooth Test Error:", get_error(predicted_smooth, y_test)
print "Saving smoothed predictions"
np.save("data/ps-test_%s" % args.type, np.array(predicted_smooth))
plot_classification(args, x_test, y_test, np.array(predicted_smooth))
def testing_from_file(ann_data_file, ann_test_data_file, dir_report):
ann = lf.neural_net()
ann.create_from_file(ann_data_file)
with open(ann_test_data_file) as f:
content = f.readlines()
count, err = len(content) // 2, 0
for i in range(1, len(content)):
if i % 2 == 1:
# input vector for ann
vec = hp.get_int_array_from_string(content[i][0:-2])
rec = hp.get_max_from_int_array(ann.run(vec))
# control vector
vec = hp.get_int_array_from_string(content[i + 1][0:-2])
num = hp.get_max_from_int_array(vec)
if num != rec:
print(rec, num)
err += 1
print "total:", count
print "error:", err
print "p={0:.3f}%".format(100 - float(err) / float(count) * 100)
def testing_from_file(ann_data_file, ann_test_data_file, dir_report):
ann = lf.neural_net()
ann.create_from_file(ann_data_file)
with open(ann_test_data_file) as f:
content = f.readlines()
count, err = len(content) // 2, 0
for i in range(1, len(content)):
if i % 2 == 1:
# input vector for ann
vec = hp.get_int_array_from_string(content[i][0:-2])
rec = hp.get_max_from_int_array(ann.run(vec))
# control vector
vec = hp.get_int_array_from_string(content[i + 1][0:-2])
num = hp.get_max_from_int_array(vec)
if num != rec:
print (rec, num)
err += 1
print "total:", count
print "error:", err
print "p={0:.3f}%".format(100 - float(err) / float(count) * 100)
def load_or_create_neural_network(path: str, input_size: int, output_size: int) -> libfann.neural_net:
if os.path.exists(path):
network = libfann.neural_net()
network.create_from_file(path)
else:
network = create_neural_network(input_size, output_size)
return network
def __init__(self, num_inputs, learning_rate, momentum, src_file=None):
"""
Constructor for a single-output neural network.
:param num_inputs: Number of inputs to the neural network.
:param learning_rate: Learning rate to use when training the neural network.
:param momentum: Learning momentum to use when training the neural network.
:param src_file: If None, then a neural network with random weights is initialized. Otherwise, the neural
network is loaded from the file.
"""
self.nn = libfann.neural_net()
if src_file is not None:
# Initialize neural network from file
self.nn.create_from_file(src_file)
else:
self.nn.create_standard_array([num_inputs, 20, 1])
self.nn.set_activation_function_hidden(libfann.SIGMOID_SYMMETRIC)
self.nn.set_activation_function_output(libfann.LINEAR)
self.nn.set_training_algorithm(libfann.TRAIN_INCREMENTAL)
self.nn.set_learning_rate(learning_rate)
self.nn.set_learning_momentum(momentum)
self.num_inputs = num_inputs
self.learning_rate = learning_rate
self.momentum = momentum
def learn(self, episodes):
state_size = len(episodes[0].states[0])
# Create the model if needed
if self._model is None:
self._model = libfann.neural_net()
self._model.create_sparse_array(
1, (state_size, self.hidden_neurons, self.nb_actions))
self._model.randomize_weights(-0.1, 0.1)
self._model.set_activation_function_layer(libfann.GAUSSIAN, 1)
self._model.set_activation_function_layer(libfann.LINEAR, 2)
# Store the values of all the states encountered in all the episodes
states = []
values = []
for episode in episodes:
states.extend(episode.states)
values.extend(episode.values)
# Train for these values
data = libfann.training_data()
data.set_train_data(states, values)
self._model.train_on_data(data, 150, 50, 1e-5)
def test(file):
"""
excepts that <filename>.net and <filename>_test.data exists
Hardkoodattu kaksi tuloa ja yksi lähtö tulostuksiin
:param file: filename without extensions.
:return:
"""
net_file = file + '.net'
data_file = file + '_test.data'
testdata = libfann.training_data() # Luo olion
testdata.read_train_from_file(data_file) # Lukee testi materiaalin joka pitäisi olla eri kuin opetusmateriaali
inputs = testdata.get_input()
outputs = testdata.get_output()
ann = libfann.neural_net()
ann.create_from_file(net_file) # Lataa aikaisemmin luotu verkko
print("MSE ERROR : %.5f" %(ann.test_data(testdata))) # Ilmoittaa verkon virheen testidatalla
for i in range(len(inputs)): # Tulostaa testidatan läpi
result = ann.run(inputs[i])
print("Input: %.2f %.2f, Output %.4f, Excepted %.4f" %(inputs[i][0], inputs[i][1], result[0], outputs[i][0] ))
def result(self):
"""
获取验证码识别结果
:return: str
"""
if self._manual:
return self._human_recognization()
neural = libfann.neural_net()
libfann.neural_net.create_from_file(neural, 'eduLogin/captcha/data/training.data')
self._binaryzation()
self._clear_noise()
image_list = self._cut_images()
captcha = ''
for image in image_list:
image = self._rotate_image(image)
image = self._resize_to_norm(image)
string = self._captcha_to_string(image)
arr = []
for x in string:
arr.append(int(x))
neural_result = libfann.neural_net.run(neural, arr)
max_element = 0
max_pos = 0
for index, item in enumerate(neural_result):
if item > max_element:
max_element = item
max_pos = index
if max_pos in range(0, 10):
captcha += str(max_pos)
else:
captcha += chr(max_pos - 10 + 97)
return captcha
def load(self, prefix):
prefix += '.' + {-1: 'l', +1: 'r'}[self.dir]
self.net = fann.neural_net()
if not self.net.create_from_file(
str(prefix + '.net')): # Must have str()
raise FileNotFoundError(str(prefix + '.net'))
self.ids.load(prefix)
def test_network(file_name):
# Grab testing file from the user
test_file = input('\nWhat file do you want to test on?: ')
# Now read from the file
file = open(test_file, 'r')
# Loop through the file line by line and grab the input and outputs
inputs = []
outputs = []
num = 0
for line in file:
array = line.split(',')
nums = [int(i) for i in array]
if num % 2 == 0:
inputs.append(nums)
else:
outputs.append(nums)
num += 1
ann = nn.neural_net()
ann.create_from_file(file_name)
# Now test the network
correct = 0
for i, inpt in enumerate(inputs):
answer = ann.run(inpt)
answer = math.ceil(outputs[i][0] - answer[0])
if answer == outputs[i][0]:
correct += 1
print('\nAccuracy is ', (correct / len(outputs)) * 100, '%\n')
def from_file(cls, name, prefix):
prefix += '.intent'
self = cls(name)
self.net = fann.neural_net()
if not self.net.create_from_file(str(prefix + '.net')): # Must have str()
raise FileNotFoundError(str(prefix + '.net'))
self.ids.load(prefix)
return self
def create_ann(self):
ann = libfann.neural_net()
ann.create_standard_array(self.layers)
ann.set_learning_rate(0.7)
ann.set_activation_function_output(libfann.SIGMOID_SYMMETRIC_STEPWISE)
ann.set_activation_function_hidden(libfann.SIGMOID_SYMMETRIC_STEPWISE)
ann.set_activation_steepness_hidden(0.4)
ann.set_activation_steepness_output(0.4)
return ann
def __init__(self, settings):
self.k = settings.k
self.networks = []
self.mses = [0]*settings.k
for i in range(0,settings.k):
self.networks.append(libfann.neural_net())
self.networks[i].create_standard_array((len(settings.parameterRanges),settings.networkSize,1))
self.networks[i].set_activation_function_output(libfann.LINEAR)
def construct(self, *args):
num_input = args[0]
num_output = args[1]
ann = libfann.neural_net()
ann.create_standard_array((num_input, self.num_hidden, num_output))
ann.set_activation_function_hidden(libfann.SIGMOID)
#ann.set_activation_function_output(libfann.SIGMOID_SYMMETRIC)
ann.set_training_algorithm(libfann.TRAIN_QUICKPROP)
return (ann)
def filter_image(ann_path, noisy_image_path, output_path):
from fann2 import libfann
from skimage import io, util
from projects.denoising.neural.filtering import filter_fann
noisy_image = util.img_as_float(io.imread(noisy_image_path))
ann = libfann.neural_net()
ann.create_from_file(ann_path)
filtered_image = filter_fann(noisy_image, ann)
io.imsave(output_path, filtered_image)
def construct(self, *args):
num_input = args[0]
num_output = args[1]
ann = libfann.neural_net()
ann.create_sparse_array(self.connection_rate,
(num_input, self.num_hidden, num_output))
ann.set_learning_rate(self.learning_rate)
ann.set_activation_function_hidden(libfann.SIGMOID_SYMMETRIC_STEPWISE)
ann.set_activation_function_output(libfann.GAUSSIAN)
return (ann)
def __init__(self, phenotype, *args, **kwargs):
self.phenotype = phenotype
self.filtered_q = None
# Initialize network
self.mlp = libfann.neural_net()
self.mlp.create_standard_array(self.phenotype.network_shape)
self.filtered_image = None
super(NeuralFilterMLP._Individual, self).__init__(*args, **kwargs)
def plot_ann_connections(result_dir):
connections = []
for filename in os.listdir(result_dir):
if not filename.endswith('.net'):
continue
set_id = int(os.path.splitext(filename)[0].split('-')[1])
filepath = os.path.join(result_dir, filename)
ann = libfann.neural_net()
ann.create_from_file(filepath)
connections.append((set_id, len(ann.get_connection_array())))
print set_id, len(ann.get_connection_array())
connections = sorted(connections, key=lambda x: x[0])
# import pdb; pdb.set_trace()
fig = plt.figure(figsize=(6, 4))
# http://stackoverflow.com/questions/3918028/how-do-i-plot-multiple-x-or-y-axes-in-matplotlib
ax = fig.add_subplot(111)
ax.spines['bottom'].set_position(('outward', 40))
make_second_bottom_spine(label='Filtravimo lango dydis')
groups = [
('3x3', (0, 13)),
('5x5', (14, 27)),
('7x7', (28, 41)),
('9x9', (42, 55)),
]
for name, xspan in groups:
annotate_group(name, xspan)
major_ticks = np.arange(0, 56, 14)
minor_ticks = np.arange(0, 56, 1)
ax.set_xticks(major_ticks)
ax.set_xticks(minor_ticks, minor=True)
# ax.set_yticks(np.arange(0, 2100, 500))
ax.grid(which='both')
# or if you want differnet settings for the grids:
ax.grid(which='minor', alpha=0.2)
ax.grid(which='major', alpha=0.5)
# plt.tight_layout()
plt.plot([
c[1] for c in connections
], label='DNT su Gauso a.f.', linewidth=2)
# plt.legend(loc=loc)
# plt.ylim([0.0, 1])
# plt.ylim([0, 2100])
plt.xlabel(u'DNT topologijos eil. nr.')
plt.ylabel(u'Jungčių skaičius')
plt.grid(True)
plt.tight_layout()
plt.show()
return connections
def _get_nn(inputs, hidden):
"""
Construct a neural network.
"""
ann = libfann.neural_net()
ann.create_standard_array((inputs, hidden[0], 1))
ann.set_learning_rate(_LEARNING_RATE)
ann.set_activation_function_hidden(libfann.SIGMOID_SYMMETRIC)
ann.set_activation_function_output(libfann.LINEAR_PIECE_SYMMETRIC)
ann.set_training_algorithm(libfann.TRAIN_INCREMENTAL)
#ann.set_rprop_delta_zero(1e-6)
return ann
def _get_nn(inputs, hidden):
"""
Construct a neural network.
"""
ann = libfann.neural_net()
ann.create_standard_array((inputs, hidden[0], 1))
ann.set_learning_rate(LEARNING_RATE)
ann.set_activation_function_hidden(libfann.SIGMOID_SYMMETRIC)
ann.set_activation_function_output(libfann.LINEAR_PIECE_SYMMETRIC)
ann.set_training_algorithm(libfann.TRAIN_RPROP)
ann.set_rprop_delta_zero(1e-6)
return ann
def _create_network(self):
""" create_sparse_array Creates a standard back propagation neural network """
self.ann = libfann.neural_net()
""" Setup the network """
self.ann.create_sparse_array(self.connection_rate, self.network)
self.ann.set_learning_rate(self.learning_rate)
self.ann.set_learning_momentum(self.learning_momentum)
self.ann.randomize_weights(-self.initial_weight, self.initial_weight)
self.ann.set_training_algorithm(self.training_algorithm)
""" set activation function """
self.ann.set_activation_function_hidden(self.hidden_activation)
self.ann.set_activation_function_output(self.output_activation)
""" This option is only used in cascading network """
def _parse_ann_info(image_res_dir):
"""
Input - directory of a single image
"""
results = {}
for result_file in os.listdir(image_res_dir):
if result_file.endswith('.net'):
ann_path = os.path.join(image_res_dir, result_file)
param_set = os.path.splitext(result_file)[0]
ann = libfann.neural_net()
ann.create_from_file(ann_path)
results[param_set] = {}
results[param_set]['connection_count'] = len(ann.get_connection_array())
return results
def fann_ble_test_recovered(self, data, test_data, normalize=True, savepath="./temp_save.conf"): if not os.path.exists(savepath): print "No File Included" return [] net = libfann.neural_net() net.create_from_file(savepath) res = zeros((test_data.mat_res.mat.shape[0], len(data.mat_res.sep_mat.keys()))) for i, test_array in enumerate(test_data.mat_res.mat): if normalize: # test_array = -test_array / 100.0 test_array = self.normalize(test_array, self.m_min, self.m_max) res[i, :] = net.run(test_array) self.net = net return res
def predict_input_button_clicked(self):
fake_id = '10000'
input_text = str(self.weibo_input_text.toPlainText()).replace(
'\n', ' ')
if input_text == '':
self.statusBar.showMessage(
'Prediction Fail! Please enter the text.', 3000)
return
input_date_time = str(
self.date_time_edit.dateTime().toString('yyyy.M.d HH:mm'))
input_forward = str(self.forward_spinbox.value())
start_time = timeit.default_timer()
input_line = fake_id + '\t' + input_text + '\t' + input_date_time + '\t' + input_forward
input_list = list()
input_list.append(input_line)
computed_list = train_prediction.process_weibo(input_list, True)
os.chdir(os.path.abspath(os.curdir + '/neural-network'))
ann = libfann.neural_net()
ann.create_from_file("trained.net")
# single_item: [0]id [1]emotion grate [2]time [3]forward [4]content
single_item = computed_list[0] # Only has the first one
result = ann.run([
float(single_item[1]),
float(single_item[2]),
float(single_item[3])
])
print('Prediction:' + str('%-18s' % result[0]))
prediction = 'No'
if (result[0] - 0.5) > 0:
prediction = 'Yes'
result_text = unicode('Content: ' + single_item[4] + '\nTime: ' +
str('%.2f' % single_item[2]) +
'(hour)\nForward: ' + single_item[3] +
'\nEmotion Grate: ' + str(single_item[1]) +
'\nSuicide prediction: ' + prediction + '\n\n')
stop_time = timeit.default_timer()
result_text += 'Total prediction time: ' + str(
'%.2f' % (stop_time - start_time)) + ' seconds.\n'
self.result_text.setText(QtCore.QString(result_text))
self.result_text.moveCursor(QtGui.QTextCursor.End)
# Back to root folder
os.chdir('..')
self.statusBar.showMessage('Prediction Complete!', 3000)
def __init__(self, retrain=False):
#####################################
# Get the qual predictions to blend #
#####################################
# Collect the ratings by the predictors
RT_T = []
for p in self.predictors:
g = pandas.read_csv(
ROOT + '/results/blend/' + p,
header=None,
names=['rating']
)['rating'].tolist()
# Convert the list to a numpy array
g = np.array(g, dtype=float)
# Subtract out the mean rating of p
# g -= np.sum(g) / len(g)
g /= 5
RT_T.append(g)
# Convert the predictor ratings into a matrix to transpose it
RT_T = np.matrix(RT_T)
RT = RT_T.T
# Get it back in list of list form, each inner list is the set of
# predictions for a single user/movie combination
self.test_inputs = RT.tolist()
##################
# Create the ANN #
##################
if retrain:
self.train()
else:
filename = ROOT + '/results/ann.net'
print("Loading ANN from file.")
print(" - path = {0}\n".format(filename))
ann = libfann.neural_net()
ann.create_from_file(filename)
self.ann = ann
print("Predicting qual ratings using the loaded ANN.")
self.predict()
def _create_network(self):
""" create_sparse_array Creates a standard back propagation neural network """
self.ann = libfann.neural_net()
""" Setup the network """
self.ann.create_sparse_array(self.connection_rate, self.network)
self.ann.set_learning_rate(self.learning_rate)
self.ann.set_learning_momentum(self.learning_momentum)
self.ann.randomize_weights(-self.initial_weight,self.initial_weight)
self.ann.set_training_algorithm(self.training_algorithm)
""" set activation function """
self.ann.set_activation_function_hidden(self.hidden_activation)
self.ann.set_activation_function_output(self.output_activation)
""" This option is only used in cascading network """
def fann_ble_test_recovered_accum(self, data, test_data, normalize=True, savepath="./temp_save.conf", accum_depth=3): if not os.path.exists(savepath): print "No File Included" return [] net = libfann.neural_net() net.create_from_file(savepath) res = zeros((test_data.mat_res.mat.shape[0], int(len(data.mat_res.sep_mat.keys())/4.0))) for i, test_array_miao in enumerate(test_data.mat_res.mat): test_array_raw = test_data.mat_res.mat[max(0, i - accum_depth):i+1, :] test_array = self.stack_data(test_array_raw) if normalize: test_array = self.normalize(test_array, self.m_min, self.m_max) # print test_array res[i, :] = net.run(test_array) self.net = net return res
def predict_file_button_clicked(self):
self.statusBar.showMessage("Opening Weibo file...")
result_text = ''
filename = QtGui.QFileDialog.getOpenFileName(self, 'Open file', './')
try:
with open(filename) as fin_predict:
start_time = timeit.default_timer()
weibo_predict_list = train_prediction.process_weibo(
fin_predict.readlines()[1:], True)
os.chdir(os.path.abspath(os.curdir + '/neural-network'))
ann = libfann.neural_net()
ann.create_from_file("trained.net")
# item: [0]id [1]emotion grate [2]time [3]forward [4]content
for index, item in enumerate(weibo_predict_list):
result = ann.run(
[float(item[1]),
float(item[2]),
float(item[3])])
print('Prediction:' + str('%-18s' % result[0]))
prediction = 'No'
if (result[0] - 0.5) > 0:
prediction = 'Yes'
result_text += unicode('Weibo ID: ' + item[0] +
'\nContent: ' + item[4] +
'\nTime: ' + str('%.2f' % item[2]) +
'(hour)\nForward: ' + item[3] +
'\nEmotion Grate: ' + str(item[1]) +
'\nSuicide prediction: ' +
prediction + '\n\n')
stop_time = timeit.default_timer()
result_text += 'Total prediction time: ' + str(
'%.2f' % (stop_time - start_time)) + ' seconds.\n'
self.result_text.setText(QtCore.QString(result_text))
self.result_text.moveCursor(QtGui.QTextCursor.End)
# Back to root folder
os.chdir('..')
self.statusBar.showMessage('Prediction Complete!', 3000)
except IOError as err:
print('Input predict file error: ' + str(err))
self.statusBar.showMessage(
"Prediction Fail! Please choose a correct file.", 3000)
def testNet(testSet, netFilename, labelHandler):
testSummaryFilename = time.strftime("testingOutput_%d_%m_%Y_%H.csv")
if NNWrapper.numThatActuallyHaveLabel == None:
NNWrapper.numThatActuallyHaveLabel = {}
NNWrapper.numThatActuallyHaveLabelCorrectlyLabeled = {}
try:
os.remove(testSummaryFilename)
except:
print "already no such file"
testingSummaryFile = open(testSummaryFilename, "a")
ann = libfann.neural_net()
ann.create_from_file(netFilename)
#ann.print_connections()
numTested = 0
stats = {}
for pair in testSet:
featureVec = pair[0]
actualLabelVec = pair[1]
result = ann.run(featureVec)
testingSummaryFile.write(str(actualLabelVec)+","+str(result)+"\n")
numTested += 1
actualLabels = labelHandler.getLabelsForXInNRep(pair[1])
guessedLabels = labelHandler.labelsFromNetAnswer(result)
for actualLabel in actualLabels:
boxStats = stats.get(actualLabel, {})
for guessedLabel in guessedLabels:
boxStats[guessedLabel] = boxStats.get(guessedLabel, 0) + 1
stats[actualLabel] = boxStats
for key in stats:
print key
print "*******************"
for label in labelHandler.labelIdsToLabels:
count = stats[key].get(label, 0)
print label, "\t\t\t", count
testingSummaryFile.close()
def read(self):
start = time.process_time()
self.client.initProtocol(self.cfg)
ann = libfann.neural_net()
ann.create_from_file("./data/net_16000.net")
proccess = processbasic.ProcessBasic.ProcessBasic(ann, self.client)
packetReader = common.PacketReader.PacketReader(proccess)
dataReader = common.DataReader.DataReader(self.cfg, packetReader)
dataReader.readDataSet()
dataReader.printFiles()
packetReader.printSamples()
proccess.printPackets()
self.client.close()
end = time.process_time()
print("Time: " + str((end - start) * 1000) + "ms")
def __init__(self, source_image, network_shape, init_method, stats, fitness_func):
self.source_image = source_image
self.target_image = None
self.network_shape = network_shape
# Statistical chromosome initialization
self.init_method = init_method
self.stats = stats
# Other
self.fitness_func = fitness_func
self.initial_q = metrics.q_py(source_image)
self.trained_anns = []
if fitness_func == 'ann':
# Instantiate all neural nets
ann_dir = os.path.join(
os.path.dirname(os.path.realpath(__file__)),
'trained_anns'
)
for filename in os.listdir(ann_dir):
if filename.endswith('.net'):
ann_path = os.path.join(ann_dir, filename)
ann = libfann.neural_net()
ann.create_from_file(ann_path)
self.trained_anns.append(ann)
else:
slope = 0.65722219398100967
intercept = 0.099529774137723237
self.ideal_q_guess = slope * self.initial_q + intercept
# print "Initial Q guess: " + str(self.ideal_q_guess)
exp_coefs = [
6.58953834,
29.54967305,
6.00895362,
-40.3269125
]
exp_p = lambda x, a, b, c, d: -a * np.exp(-b * x + c) + d
exp = lambda x: exp_p(x, *exp_coefs)
self.parabola_coef = exp(self.ideal_q_guess)
s = "INITIAL Q: " + str(self.initial_q)
s += ", IDEAL Q GUESS: " + str(self.ideal_q_guess)
s += ", PARABOLA COEF: " + str(self.parabola_coef)
print s
def build_pheno(self, p, pop):
ann = libfann.neural_net()
#ann.set_activation_function_output (libfann.SIGMOID_SYMMETRIC_STEPWISE)
#ann.set_activation_function_hidden (libfann.SIGMOID_SYMMETRIC_STEPWISE)
# connection_rate, (inputs, hidden, outputs)
ann.create_sparse_array(1, (2, 2, 1))
# get_num_layers get_num_input get_num_output
# get_total_connections get_total_neurons
# init_weights (self, data) print_connections, print_parameters
# set_weight (self, from_neuron, to_neuron, weight)
for frm in range(3):
for to in range(2):
v = self.get_allele(p, pop, frm * 2 + to)
ann.set_weight(frm, to + 3, v)
#print (frm, to + 3, ann.get_weight (frm, to + 3))
for frm in range(3):
ann.set_weight(frm + 3, 6, self.get_allele(p, pop, frm + 6))
#print (frm + 3, 6, ann.get_weight (frm + 3, 6))
return ann
def testNet(testSet, netFilename, labelHandler):
if NNWrapper.numThatActuallyHaveLabel == None:
NNWrapper.numThatActuallyHaveLabel = {}
NNWrapper.numThatActuallyHaveLabelCorrectlyLabeled = {}
try:
os.remove(testingSummaryFilename)
except:
print "already no such file"
testingSummaryFile = open(NNWrapper.testingSummaryFilename, "a")
ann = libfann.neural_net()
ann.create_from_file(netFilename)
#ann.print_connections()
numTested = 0
numLabeledCorrectly = 0
for pair in testSet:
featureVec = pair[0]
actualLabel = labelHandler.getLabelForOneInNRep(pair[1])
result = ann.run(featureVec)
#print result, actualLabel
numTested += 1
NNWrapper.numThatActuallyHaveLabel[actualLabel] = NNWrapper.numThatActuallyHaveLabel.get(actualLabel, 0) + 1
guessedLabel = labelHandler.closestLabel(result)
testingSummaryFile.write(guessedLabel+","+actualLabel+"\n")
if actualLabel == guessedLabel:
numLabeledCorrectly += 1
NNWrapper.numThatActuallyHaveLabelCorrectlyLabeled[actualLabel] = NNWrapper.numThatActuallyHaveLabelCorrectlyLabeled.get(actualLabel, 0) + 1
print "numTested", numTested
print "numLabeledCorrectly", numLabeledCorrectly
NNWrapper.totalTested += numTested
NNWrapper.totalCorrect += numLabeledCorrectly
print "totalTested", NNWrapper.totalTested
print "totalCorrect", NNWrapper.totalCorrect
print "percentageCorrect", float(NNWrapper.totalCorrect)/NNWrapper.totalTested
print "*****************"
for key in NNWrapper.numThatActuallyHaveLabel:
print key, NNWrapper.numThatActuallyHaveLabel.get(key,0), NNWrapper.numThatActuallyHaveLabelCorrectlyLabeled.get(key,0), float(NNWrapper.numThatActuallyHaveLabelCorrectlyLabeled.get(key,0))/NNWrapper.numThatActuallyHaveLabel.get(key,0)
testingSummaryFile.close()
def train(self,
training_file_path,
num_inputs,
num_outputs,
nn_path=DEFAULT_ANN_PATH,
num_hid_neurons=None):
"""
Trains an ANN from data containing in a text-based training file that is created by xml2trainingdata.py.
:param two_hid: specifies whether to use two hidden layers or not (default not)
:param training_file_path: path to training file
:param nn_path: path to save ANN to
:param num_inputs: number of input neurons
:param num_outputs: number of output neurons
:param num_hid_neurons: number of hidden neurons
:return: None
"""
# if hidden neurons are not specified, set the number to 2/3 of the sum of input and output neurons
if num_hid_neurons is None:
num_hid_neurons = (2 * (num_inputs * num_outputs)) / 3
ann_tuple = (num_inputs, num_hid_neurons, num_outputs)
# create the ANN
ann = fann.neural_net()
ann.create_sparse_array(CONNECTION_RATE, ann_tuple)
# set learning style
ann.set_learning_rate(LEARNING_RATE)
# set activation function
ann.set_activation_function_output(fann.SIGMOID_SYMMETRIC)
# train the ANN on file
ann.train_on_file(training_file_path, MAX_ITERATIONS,
ITERATIONS_BETWEEN_REPORTS, DESIRED_ERROR)
# save ann to file and free memory associated with it
ann.save(nn_path)
ann.destroy()
# set own ann to saved ann
self.load_ann_from_file(nn_path)
def main():
args = parse_args()
print "Loading data"
x_train = np.load("data/x-train-pca%s_%s.npy" % (args.pca, args.type))
x_valid = np.load("data/x-valid-pca%s_%s.npy" % (args.pca, args.type))
y_train = np.load("data/y-train_%s.npy" % args.type)
y_valid = np.load("data/y-valid_%s.npy" % args.type)
print "x_train Shape", x_train.shape
print "x_valid Shape", x_valid.shape
print "y_train Shape", y_train.shape
print "y_valid Shape", y_valid.shape
print "Creating ann"
ann = libfann.neural_net()
ann.create_standard_array((x_train.shape[1], num_neurons_hidden, y_train.shape[1]))
ann.set_bit_fail_limit(0.5)
ann.set_learning_rate(learning_rate)
print "Converting training data"
train_data = XY_to_fann_train_data(x_train, y_train)
print "Converting testing data"
test_data = XY_to_fann_train_data(x_valid, y_valid)
ann.train_on_data(train_data, max_epochs, 1, desired_error)
ann.test_data(train_data)
bit_fail = ann.get_bit_fail()
train_err = (float(bit_fail) / (y_train.shape[0] * y_train.shape[1]))
print "Train Error:", train_err
ann.test_data(test_data)
bit_fail = ann.get_bit_fail()
valid_err = (float(bit_fail) / (y_valid.shape[0] * y_valid.shape[1]))
print "Validation Error:", valid_err
print "Saving ANN to %s" % args.fann_file
ann.save("nnets/%s" % args.fann_file)
def fann_ble_test_recovered(self,
data,
test_data,
normalize=True,
savepath="./temp_save.conf"):
if not os.path.exists(savepath):
print "No File Included"
return []
net = libfann.neural_net()
net.create_from_file(savepath)
res = zeros(
(test_data.mat_res.mat.shape[0], len(data.mat_res.sep_mat.keys())))
for i, test_array in enumerate(test_data.mat_res.mat):
if normalize:
# test_array = -test_array / 100.0
test_array = self.normalize(test_array, self.m_min, self.m_max)
res[i, :] = net.run(test_array)
self.net = net
return res
def __init__(self):
super().__init__()
self.learnbuffer = []
self.dumbtraining = False
connection_rate = 1
num_input = 2
#hidden = (40,40)
hidden = (50,)
#hidden = (20,10,7)
num_output = 4
learning_rate = 0.7
self.NN = libfann.neural_net()
#self.NN.set_training_algorithm(libfann.TRAIN_BATCH)
#self.NN.set_training_algorithm(libfann.TRAIN_RPROP)
#self.NN.set_training_algorithm(libfann.TRAIN_QUICKPROP)
self.NN.create_sparse_array(connection_rate, (num_input,)+hidden+(num_output,))
self.NN.randomize_weights(-1,1)
self.NN.set_learning_rate(learning_rate)
self.NN.set_activation_function_hidden(libfann.SIGMOID_SYMMETRIC_STEPWISE)
self.NN.set_activation_function_output(libfann.SIGMOID_SYMMETRIC_STEPWISE)
def entrenar():
connection_rate = 1
learning_rate = 0.7
num_input = 400
num_hidden = 50
num_output = 5
desired_error = 0.0001
max_iterations = 100000
iterations_between_reports = 1000
ann = libfann.neural_net()
ann.create_sparse_array(connection_rate, (num_input, num_hidden, num_output))
ann.set_learning_rate(learning_rate)
ann.set_activation_function_output(libfann.SIGMOID_SYMMETRIC_STEPWISE)
ann.set_activation_function_hidden(libfann.SIGMOID_SYMMETRIC_STEPWISE)
ann.train_on_file("patrones.data", max_iterations, iterations_between_reports, desired_error)
ann.save("red.net")
def trainNetwork(dataFilename, netFilename, layerSizes, max_iterations, desired_error): # layerSizes should look something like this: (numInput, 200, 80, 40, 20, 10, numOutput) ann = libfann.neural_net() #ann.create_sparse_array(NNWrapper.connection_rate, (numInput, 6, 4, numOutput)) #TODO: is this what we want? # the one that works in 40 seconds 4, 10, 6, 1. the one that trained in 30 secs was 6,6 ann.create_sparse_array(NNWrapper.connection_rate, layerSizes) ann.set_learning_rate(NNWrapper.learning_rate) ann.set_activation_function_output(libfann.SIGMOID_SYMMETRIC_STEPWISE) ann.set_bit_fail_limit(.1) #ann.randomize_weights(0,0) t0 = time.clock() ann.train_on_file(dataFilename, max_iterations, NNWrapper.iterations_between_reports, desired_error) t1 = time.clock() seconds = t1-t0 m, s = divmod(seconds, 60) h, m = divmod(m, 60) print "Time to train:" print "%d:%02d:%02d" % (h, m, s) ann.save(netFilename)
def train_network(file):
print('\nTraining Network from this file:', file, '\n')
# Setup all the variables
connection_rate = 1
learning_rate = 0.7
num_input = 2
num_hidden = 4
num_output = 1
desired_error = 0.0001
max_iterations = 100000
iterations_between_reports = 1000
ann = nn.neural_net()
ann.create_sparse_array(connection_rate, (num_input, num_hidden, num_output))
ann.set_learning_rate(learning_rate)
ann.set_activation_function_output(nn.SIGMOID_SYMMETRIC_STEPWISE)
ann.train_on_file(file, max_iterations, iterations_between_reports, desired_error)
ann.save(file.replace('.data', '.net'))
def predict_input_button_clicked(self):
fake_id = '10000'
input_text = str(self.weibo_input_text.toPlainText()).replace('\n', ' ')
if input_text == '':
self.statusBar.showMessage('Prediction Fail! Please enter the text.', 3000)
return
input_date_time = str(self.date_time_edit.dateTime().toString('yyyy.M.d HH:mm'))
input_forward = str(self.forward_spinbox.value())
start_time = timeit.default_timer()
input_line = fake_id + '\t' + input_text + '\t' + input_date_time + '\t' + input_forward
input_list = list()
input_list.append(input_line)
computed_list = train_prediction.process_weibo(input_list, True)
os.chdir(os.path.abspath(os.curdir + '/neural-network'))
ann = libfann.neural_net()
ann.create_from_file("trained.net")
# single_item: [0]id [1]emotion grate [2]time [3]forward [4]content
single_item = computed_list[0] # Only has the first one
result = ann.run([float(single_item[1]), float(single_item[2]), float(single_item[3])])
print('Prediction:' + str('%-18s' % result[0]))
prediction = 'No'
if (result[0] - 0.5) > 0:
prediction = 'Yes'
result_text = unicode('Content: ' + single_item[4] + '\nTime: ' + str('%.2f' % single_item[2]) +
'(hour)\nForward: ' + single_item[3] + '\nEmotion Grate: ' +
str(single_item[1]) + '\nSuicide prediction: ' + prediction + '\n\n')
stop_time = timeit.default_timer()
result_text += 'Total prediction time: ' + str('%.2f' % (stop_time - start_time)) + ' seconds.\n'
self.result_text.setText(QtCore.QString(result_text))
self.result_text.moveCursor(QtGui.QTextCursor.End)
# Back to root folder
os.chdir('..')
self.statusBar.showMessage('Prediction Complete!', 3000)
def run_fann(dataFile):
with open(dataFile) as f:
connection_rate = 1
learning_rate = 0.7
num_hidden = 10
_, num_input, num_output = tuple(map(lambda d: int(d),
f.readline().split()))
desired_error = 0.0001
max_iterations = 10000
iterations_between_reports = 1000
ann = libfann.neural_net()
ann.create_sparse_array(connection_rate,
(num_input, num_hidden, num_output))
ann.set_learning_rate(learning_rate)
ann.set_activation_function_output(libfann.SIGMOID_SYMMETRIC_STEPWISE)
ann.train_on_file(dataFile, max_iterations,
iterations_between_reports, desired_error)
ann.save(os.path.splitext(dataFile)[0] + ".net")
def learn(train_file):
"""
:param train_file: filename without extensions.
:return:
method will create nn called <filename>.net
and except that <filename>.data exists
Kaikki arvot on kovakoodattu tällä hetkellä. Voisi olla vapaaehtoisia parametrejä.
Hidden layerien määrän hallintaan pitää keksiä jokin juttu.
"""
net_file = train_file + '.net'
data_file = train_file + '_train.data'
connection_rate = 1
learning_rate = 0.5 # learning rate ei saa olla liian suuri, toisaalta liian pienellä oppiminen kestää kauan
num_input = 2
num_hidden = 4
num_output = 1
desired_error = 0.00005
max_iterations = 100000
iterations_between_reports = 10
ann = libfann.neural_net()
# Lue tiedosto
trainindata = libfann.training_data()
trainindata.read_train_from_file(data_file)
# Luo verkon
ann.create_sparse_array(connection_rate, (num_input, num_hidden, num_hidden, num_output))
ann.set_learning_rate(learning_rate)
ann.set_activation_function_output(libfann.SIGMOID_SYMMETRIC_STEPWISE) # Aktivointi funktio
ann.set_training_algorithm(libfann.TRAIN_INCREMENTAL) # Oppimis algoritmi
ann.train_on_data(trainindata, max_iterations, iterations_between_reports, desired_error)
ann.save(net_file)
def fann_ble_test_recovered_accum(self,
data,
test_data,
normalize=True,
savepath="./temp_save.conf",
accum_depth=3):
if not os.path.exists(savepath):
print "No File Included"
return []
net = libfann.neural_net()
net.create_from_file(savepath)
res = zeros((test_data.mat_res.mat.shape[0],
int(len(data.mat_res.sep_mat.keys()) / 4.0)))
for i, test_array_miao in enumerate(test_data.mat_res.mat):
test_array_raw = test_data.mat_res.mat[max(0, i - accum_depth):i +
1, :]
test_array = self.stack_data(test_array_raw)
if normalize:
test_array = self.normalize(test_array, self.m_min, self.m_max)
# print test_array
res[i, :] = net.run(test_array)
self.net = net
return res
def trainANN(features, labels, connRate, hidNodes, error, binary):
"""
Train the neural network using the given training data and
parameters. Returns a fully trained ANN.
"""
# Organize ANN parameters
connection_rate = connRate
num_input = 72
num_hidden = hidNodes
num_output = 3
desired_error = error
max_iterations = 100000
# Print out two reports for every ANN
iterations_between_reports = 50000
# Binarize labels as it is necessary for ANN
labels = binary.fit_transform(labels)
# Cast numpy to python list
annFeatures = features.tolist()
annLabels = labels.tolist()
# Create an ANN training data instance and set data
training = libfann.training_data()
training.set_train_data(annFeatures, annLabels)
ann = libfann.neural_net()
ann.create_sparse_array(connection_rate,
(num_input, num_hidden, num_output))
# Train the ANN
ann.train_on_data(training, max_iterations, iterations_between_reports,
desired_error)
return ann
def __init__(self):
super().__init__()
self.learnbuffer = []
self.dumbtraining = False
connection_rate = 1
num_input = 2
#hidden = (40,40)
hidden = (50, )
#hidden = (20,10,7)
num_output = 4
learning_rate = 0.7
self.NN = libfann.neural_net()
#self.NN.set_training_algorithm(libfann.TRAIN_BATCH)
#self.NN.set_training_algorithm(libfann.TRAIN_RPROP)
#self.NN.set_training_algorithm(libfann.TRAIN_QUICKPROP)
self.NN.create_sparse_array(connection_rate,
(num_input, ) + hidden + (num_output, ))
self.NN.randomize_weights(-1, 1)
self.NN.set_learning_rate(learning_rate)
self.NN.set_activation_function_hidden(
libfann.SIGMOID_SYMMETRIC_STEPWISE)
self.NN.set_activation_function_output(
libfann.SIGMOID_SYMMETRIC_STEPWISE)
def train(captchas_dir):
NUM_INPUT = CH_WIDTH * CH_HEIGHT
NUM_NEURONS_HIDDEN = NUM_INPUT // 3
NUM_OUTPUT = 10
ann = libfann.neural_net()
ann.create_standard_array((NUM_INPUT, NUM_NEURONS_HIDDEN, NUM_OUTPUT))
# ann.set_activation_function_hidden(libfann.SIGMOID)
# ann.set_activation_function_output(libfann.SIGMOID)
# ann.randomize_weights(0.0, 0.0)
start = time.time()
succeed = 0
captchas_dir = os.path.abspath(captchas_dir)
captchas = os.listdir(captchas_dir)
report()
for i, name in enumerate(captchas):
answer = re.match(r'(\d{6})\.png$', name)
if not answer:
continue
answer = answer.group(1)
fpath = os.path.join(captchas_dir, name)
try:
img = get_image(fpath)
ch_imgs = segment(img)
for ch_img, digit in zip(ch_imgs, answer):
ann.train(get_ch_data(ch_img), make_ann_output(digit))
except Exception as exc:
report('Error occured while processing {}: {}'.format(name, exc))
report()
else:
succeed += 1
report('{}/{}'.format(i + 1, len(captchas)), progress=True)
runtime = time.time() - start
report('Done training on {}/{} captchas in {:.3f} seconds'.format(
succeed, len(captchas), runtime))
return ann
def test_prediction(test_list):
os.chdir(os.path.abspath(os.curdir + '/neural-network'))
ann = libfann.neural_net()
ann.create_from_file("trained.net")
correct_count = 0
# item: [0]id [1]emotion grate [2]time [3]forward [4]suicide or not
for item in test_list:
result = ann.run([float(item[1]), float(item[2]), float(item[3])])
print('Prediction:' + str('%-18s' % result[0]) + 'Suicide:' + item[4])
prediction = 0
if (result[0] - 0.5) > 0:
prediction = 1
if int(item[4]) == prediction:
correct_count += 1
accuracy = round(float(correct_count) / len(test_list) * 100, 2)
print('Results: Correct number is ' + str(correct_count) + '. Model accuracy is ' + str('%.2f' % accuracy) + '%.')
# Back to root folder
os.chdir('..')
return accuracy