def op2():
"""
generate rotation data
:return:
"""
data = common.Data("mnist/mnist_train/train_data.npy",
"mnist/mnist_train/mnist_train_label",
"mnist/mnist_test/test_data.npy",
"mnist/mnist_test/mnist_test_label", 1, 28)
fig = 28
d = []
l = []
for i in range(len(data.train_x)):
data.train_x[i] = data.train_x[i].astype(np.uint8)
img = Image.fromarray(data.train_x[i].reshape(fig,
fig).astype(np.uint8))
img_r = img.rotate(45)
img_l = img.rotate(315)
matrix = np.asarray(img_r)
matrix = matrix.reshape(fig * fig, )
d.append(data.train_x[i])
d.append(matrix)
matrix = np.asarray(img_l)
matrix = matrix.reshape(fig * fig, )
d.append(matrix)
l.append(data.train_y_no_one_hot[i])
l.append(data.train_y_no_one_hot[i])
l.append(data.train_y_no_one_hot[i])
np.save('mnist/mnist_train/rotate_train.npy', d)
np.save('mnist/mnist_train/rotate_label.npy', l)
def main():
data = common.Data(common_path+"/mnist_train/train_data.npy", common_path+"/mnist_train/mnist_train_label",
common_path+"/mnist_test/test_data.npy", common_path+"/mnist_test/mnist_test_label", fig)
train_x = data.train_x
test_x = data.test_x
train_y = data.train_y
test_y = data.test_y
SVM(train_x, test_x, train_y, test_y,('rbf',1.3))
def main():
common.configure_logging('korred')
data = common.Data()
handler = Handler(data=data)
NativeMessageInterface(
callback=handler.handle,
interactive=os.getenv('INTERACTIVE', False),
).run()
def main():
data = common.Data(common_path + "/mnist_train/train_data.npy",
common_path + "/mnist_train/mnist_train_label",
common_path + "/mnist_test/test_data.npy",
common_path + "/mnist_test/mnist_test_label", fig)
train_x = data.train_x
test_x = data.test_x
train_y = data.train_y
test_y = data.test_y
Autoencoder(train_x, test_x, train_y, test_y)
def op0():
"""
generate lowest 15000 confident images
:return: none
"""
data = common.Data("mnist/mnist_train/train_data.npy",
"mnist/mnist_train/mnist_train_label",
"mnist/mnist_test/test_data.npy",
"mnist/mnist_test/mnist_test_label", 1, 28)
res = common.predict('model/1.4.0', 60000, data.train_x, 28)
common.gen_data(res, data.train_x, data.train_y_no_one_hot, 15000)
def main():
data = common.Data()
with Daemon(data):
# TODO: Check for configuration validity and ask user if the old configuration should be kept
NativeMessagingConfiguration().write()
launchctl_manager = LaunchctlManager()
app = App(launchctl_manager)
menu_item = rumps.MenuItem(title='Launch at Login')
menu_item.state = launchctl_manager.is_loaded()
app.menu = [
'Install Firefox Extension...',
menu_item,
'View Logs',
]
app.run()
def op1():
"""
generate fc2 output for svm
:return: none
"""
data = common.Data("mnist/mnist_train/train_data.npy",
"mnist/mnist_train/mnist_train_label",
"mnist/mnist_test/test_data.npy",
"mnist/mnist_test/mnist_test_label", 1, 28)
res = common.predict('CNN/model/SVM1', 60000, data.test_x, 28, "out1")
data_fc = []
for i in range(len(res)):
data_fc.append(res[i][0][0][0])
data_fc = np.array(data_fc)
np.save('mnist/mnist_test/fc1_5.npy', data_fc)
Reparameterization trick by sampling fr an isotropic unit Gaussian.
:param z_mean: mean of Gaussian variable z
:param z_log_var: covariance matrix of Gaussian variable z
:return z sampled from z_mean and z_log_var
"""
batch = K.shape(z_mean)[0]
dim = K.int_shape(z_mean)[1]
# by default, random_normal has mean=0 and std=1.0
epsilon = K.random_normal(shape=(batch, dim))
return z_mean + K.exp(0.5 * z_log_var) * epsilon
# MNIST dataset
data = common.Data(common_path + "/mnist_train/train_data.npy",
common_path + "/mnist_train/mnist_train_label",
common_path + "/mnist_test/test_data.npy",
common_path + "/mnist_test/mnist_test_label", fig)
x_train = data.train_x
x_test = data.test_x
y_train = data.train_y
y_test = data.test_y
image_size = x_train.shape[1]
original_dim = image_size * image_size
x_train = np.reshape(x_train, [-1, original_dim])
x_test = np.reshape(x_test, [-1, original_dim])
# network parameters
input_shape = (original_dim, )
intermediate_dim = 512
batch_size = 128
# print(x.shape)
# print(self.line.weight.shape)
nn.init.normal_(self.out.weight)
def forward(self, x):
re = self.line(x)
re = self.active(re)
re = self.line2(re)
re = self.active2(re)
re = self.line3(re)
re = self.active3(re)
re = self.out(re)
return re
dataset = c.Data(x, y)
dataLoader = tud.DataLoader(dataset, batch_size=20, shuffle=True)
model = NetWork3()
loss_fn = nn.MSELoss()
optimer = torch.optim.Adam(model.parameters(), lr=1e-4)
# schuler = torch.optim.lr_scheduler.ExponentialLR(optimer, 0.5)
loss_arr = []
def train(epochs):
time = 0
last_loss = -1
for epoch in range(1, epochs + 1):
model.train()
losses = []
"""
Read data from CNN_SVM
"""
import common
import numpy as np
from sklearn.preprocessing import scale
from sklearn.svm import SVC
from sklearn.metrics import classification_report
from sklearn.metrics import accuracy_score
data = common.Data("../mnist/mnist_train/train_data.npy",
"../mnist/mnist_train/mnist_train_label",
"../mnist/mnist_test/test_data.npy",
"../mnist/mnist_test/mnist_test_label", 1, 28)
train = np.load('../mnist/mnist_train/fc1_5.npy')
test = np.load('../mnist/mnist_test/fc1_5.npy')
train = scale(train)
test = scale(test)
clf = SVC(kernel='rbf')
clf.fit(train, data.train_y_no_one_hot)
y_pred = clf.predict(test)
print(classification_report(data.test_y_no_one_hot, y_pred))
print(accuracy_score(data.test_y_no_one_hot, y_pred))
import common
"""
Variable Definition
batch_size: batch size
fig: image size
max_epoch: max iteration
common_path = common path of input data
"""
n = 5
batch_size = 500
fig = 45
max_epoch = 200
common_path = "../mnist"
data = common.Data(common_path + "/mnist_train/mnist_train_data",
common_path + "/mnist_train/mnist_train_label",
common_path + "/mnist_test/mnist_test_data",
common_path + "/mnist_test/mnist_test_label", 1, fig)
X = data.train_x.reshape(data.size, fig, fig, 1)
Y = data.train_y
testX = data.test_x.reshape(data.size_test, fig, fig, 1)
testY = data.test_y
img_prep = tflearn.ImagePreprocessing()
img_prep.add_featurewise_zero_center(per_channel=True)
img_aug = tflearn.ImageAugmentation()
net = tflearn.input_data(shape=[None, 45, 45, 1],
data_preprocessing=img_prep,
data_augmentation=img_aug)
common_path = "../mnist"
if os.path.exists(log_dir):
os.remove(log_dir)
if os.path.exists(model_dir):
os.remove(model_dir)
os.makedirs(log_dir, exist_ok=True)
os.makedirs(model_dir, exist_ok=True)
starttime = datetime.datetime.now()
logger = common.create_logger(
'CNN', log_format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
data = common.Data(common_path + "/mnist_train/rotate_train.npy",
common_path + "/mnist_train/rotate_label.npy",
common_path + "/mnist_test/test_data.npy",
common_path + "/mnist_test/mnist_test_label", batch_size,
fig)
def weight_variable(shape, name):
initial = tf.truncated_normal(shape, stddev=0.1)
return tf.Variable(initial, name=name)
def bias_variable(shape, name):
initial = tf.constant(0.1, shape=shape)
return tf.Variable(initial, name=name)
def conv2d(x, W):
