#from sklearn import svm
from sklearn.svm import SVC
from sklearn.preprocessing import MinMaxScaler
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression
from sklearn.tree import DecisionTreeClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.preprocessing import StandardScaler
from sklearn.decomposition import PCA
from sklearn.pipeline import Pipeline
import time
print('reading data...', end='')
X, y, classes = read_mnist('mnist_public.npy')
print(X) # X is the image points
print(y) # y is the answers
print('done!')
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=0.1,
train_size=0.9,
random_state=5)
scl = MinMaxScaler()
pca = PCA()
knn = KNeighborsClassifier(n_neighbors=3,
weights='uniform',
algorithm='brute',
images = tf.layers.flatten(images)
hidden = tf.layers.dense(images, 512, tf.nn.elu)
mean = tf.layers.dense(hidden, code_size, tf.nn.elu)
return mean
def make_decoder(code, data_shape=[28, 28, 1]):
hidden = tf.layers.dense(code, 512, tf.nn.elu)
logit = tf.layers.dense(hidden, np.prod(data_shape), tf.nn.elu)
logit = tf.reshape(logit, [-1] + data_shape)
return logit
if __name__ == '__main__':
mnist_folder = os.path.join('..', 'data', 'mnist.npz')
(train_x, train_y), (_, _) = mnist_utils.read_mnist(mnist_folder, flatten=False)
train_y = np.argmax(train_y, axis=1)
train_x_0 = train_x
# np.random.shuffle(train_x_0)
train_x_0 = train_x_0[:min(len(train_x_0), 1000)]
code_size = 100
tf.reset_default_graph()
inputs = tf.placeholder(tf.float32, [None, 28, 28, 1])
encoder = make_encoder(inputs, code_size)
output = make_decoder(encoder)
loss = tf.losses.huber_loss(inputs, output)
optimize = tf.train.AdamOptimizer(learning_rate=0.0001).minimize(loss)
mnist_utils.safe_mkdir('checkpoints')