def get_data():
((train_image, train_label),
(test_image,
test_label)) = load_mnist_dataset.load_mnist_dataset(one_hot=False,
normarize=True,
flatten=True)
return test_image, test_label
# official
import sys
import os
import numpy as np
from PIL import Image
# made by me
import load_mnist_dataset
sys.path.append(os.pardir)
def show_img(img):
pil_img = Image.fromarray(img)
pil_img.show()
((x_train, t_train), (x_test, t_test)) = load_mnist_dataset.load_mnist_dataset(one_hot=False, normarize=False, flatten=True)
img = x_train[0]
label = t_train[0]
print(label)
print(img.shape)
img = img.reshape((28, 28))
print(img.shape)
show_img(img)
import numpy as np
from load_mnist_dataset import load_mnist_dataset
from twolayernet import TwoLayersNet
import matplotlib.pyplot as plt
(train_images, train_label), (test_images, test_label) =\
load_mnist_dataset(flatten=False)
train_loss_list = []
iters_num = 10000
train_size = train_images.shape[0]
batch_size = 100
learning_rate = 0.1
network = TwoLayersNet(input_size=784, hidden_size=100, output_size=10)
for i in range(iters_num):
batch_mask = np.random.choice(train_size, batch_size)
image_batch = train_images[batch_mask]
label_batch = train_label[batch_mask]
grad = network.numerical_gradient(image_batch, label_batch)
for key in ("w1", "b1", "w2", "b2"):
network.parameters[key] -= grad[key] * learning_rate
loss = network.loss(image_batch, label_batch)
train_loss_list.append(loss)
def random_batch_choice(x, y, batch_size):
data_size = x.shape[0]
batch_filter = np.random.choice(data_size, batch_size)
x_batch = x[batch_filter]
y_batch = y[batch_filter]
return x_batch, y_batch
if __name__ == "__main__":
((train_images, train_labels),
(test_images, test_labels)) = load_mnist_dataset.load_mnist_dataset()
print("Non-batch")
print(train_images.shape)
print(train_labels.shape)
print(test_images.shape)
print(test_labels.shape)
(x_batch, y_batch) = random_batch_choice(train_labels,
train_labels,
batch_size=100)
print("batch")
print(x_batch.shape)
print(y_batch.shape)
#def load_mnist():
#
# # 学習データ
# x_train = np.load('/root/userspace/public/lesson1/data/x_train.npy')
# y_train = np.load('/root/userspace/public/lesson1/data/y_train.npy')
#
# テストデータ
# x_test = np.load('/root/userspace/public/lesson1/data/x_test.npy')
# x_train = x_train.reshape(-1, 784).astype('float32') / 255
# x_test = x_test.reshape(-1, 784).astype('float32') / 255
# y_train = np.eye(10)[y_train.astype('int32').flatten()]
# return (x_train, x_test, y_train)
(x_train, y_train), (x_test, y_test) = load_mnist_dataset()
model = Sequential()
model.add(
Dense(512,
input_shape=(784, ),
activation='relu',
kernel_initializer='he_normal'))
model.add(Dropout(0.2))
model.add(
Dense(256,
activation='relu',
kernel_initializer='he_normal',
kernel_regularizer=None))
model.add(Dropout(0.5))