如何使用 Keras 创建 MNIST？

Question

我开始学习深度学习，但遇到了问题。我从 Andrew W. Trask 的 Grokking Deep Learning 一书开始我的教育。我目前正在学习第 8 章（创建我自己的 MNIST 网络）。没关系，我做到了。但我想了解如何使用 DL 库，如 keras、tentherflow 等。我找到了tutorial，但它使用不同的网络。我有兴趣用 keras 克隆我的网络。我写了代码，但它不能正常工作，我不知道错误在哪里。所以我需要帮助:)

没有 Keras 的代码

import sys, numpy as np
from tensorflow.keras.datasets import mnist

(x_train, y_train), (x_test, y_test) = mnist.load_data()
images, labels = (x_train[0:1000].reshape(1000, 28*28) / 255, y_train[0:1000])

one_hot_labels = np.zeros((len(labels), 10))

for i, l in enumerate(labels):
    one_hot_labels[i][l] = 1
labels = one_hot_labels

test_images = x_test.reshape(len(x_test), 28*28) / 255
test_labels = np.zeros((len(y_test), 10))
for i, l in enumerate(y_test):
    test_labels[i][l] = 1

np.random.seed(1)
relu = lambda x: (x>=0) * x
relu2deriv = lambda x: x >= 0

batch_size = 100
alpha, iterations = (0.001, 300)
pixel_per_image, num_labels, hidden_size = (784, 10, 100)

weights_0_1 = 0.2 * np.random.random((pixel_per_image, hidden_size)) - 0.1
weights_1_2 = 0.2 * np.random.random((hidden_size, num_labels)) - 0.1

for j in range(iterations):
    error, correct_cnt = (0.0, 0)
    
    for i in range(int(len(images) / batch_size)):
        batch_start, batch_end = ((i * batch_size), ((i+1) * batch_size))

        layer_0 = images[batch_start:batch_end]

        layer_1 = relu(np.dot(layer_0,weights_0_1))
        dropout_mask = np.random.randint(2, size=layer_1.shape)
        layer_1 *= dropout_mask * 2

        layer_2 = np.dot(layer_1,weights_1_2)

        error += np.sum((labels[batch_start:batch_end] - layer_2) ** 2)

        for k in range(batch_size):
            correct_cnt += int(np.argmax(layer_2[k:k+1]) == np.argmax(labels[batch_start+k:batch_start+k+1]))

            layer_2_delta = (labels[batch_start:batch_end] - layer_2) / batch_size
            layer_1_delta = layer_2_delta.dot(weights_1_2.T) * relu2deriv(layer_1)
            layer_1_delta *= dropout_mask
            weights_1_2 += alpha * layer_1.T.dot(layer_2_delta)
            weights_0_1 += alpha * layer_0.T.dot(layer_1_delta)

    sys.stdout.write(f"\r I:{str(j)} Train-Err:{str(error/float(len(images)))[0:5]} Train-Acc:{str(correct_cnt/float(len(images)))}")

error, correct_cnt = (0.0, 0)
print(len(test_images))
for i in range(len(test_images)):
    layer_0 = test_images[i:i+1]
    layer_1 = relu(np.dot(layer_0,weights_0_1))
    layer_2 = np.dot(layer_1,weights_1_2)

    error += np.sum((test_labels[i:i+1] - layer_2) ** 2)
    correct_cnt += int(np.argmax(layer_2) == np.argmax(test_labels[i:i+1]))
sys.stdout.write(f" Test-Err:{str(error/float(len(test_images)))[0:5]} Test-Acc:{str(correct_cnt/float(len(test_images)))}")
print()

结果：Test-Err：0.428 Test-Acc：0.7992

我的 Keras 代码

import numpy as np
from tensorflow import keras
from tensorflow.keras import layers
from keras.optimizers import RMSprop 

(x_train, y_train), (x_test, y_test) = keras.datasets.mnist.load_data()

# Scale images to the [0, 1] range
x_train = x_train[0:1000].reshape(1000, 28*28).astype('float32')  / 255
y_train = y_train[0:1000]
x_test = x_test.reshape(len(x_test), 28*28).astype('float32')  / 255
print("x_train shape:", x_train.shape)
print(x_train.shape[0], "train samples")
print(x_test.shape[0], "test samples")


# convert class vectors to binary class matrices
y_train = keras.utils.to_categorical(y_train, 10)
y_test = keras.utils.to_categorical(y_test, 10)

model = keras.Sequential()
model.add(layers.Dense(100, input_shape=(784,), activation='relu'))
model.add(layers.Dropout(0.5))
model.add(layers.Dense(10))
model.summary()

batch_size = 100
epochs = 300

model.compile(loss = 'categorical_crossentropy', 
   optimizer = RMSprop(), 
   metrics = ['accuracy'])
model.fit(x_train, y_train, batch_size=batch_size, epochs=epochs)

结果：loss：8.0127 - accuracy：0.1140

很奇怪。而且 Keras 代码比我的快很多倍。有点奇怪:)