MacauAIChallenge2019_Lessons_2

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
#使用keras作為我們的語法並import有用的function  
import keras
from keras import models, layers   
from keras.datasets import mnist    
from keras.models import Sequential    
from keras.layers import Dense, Dropout, LeakyReLU, Conv2D
from keras.optimizers import RMSprop    #optimizer可選擇rmsprop, adadelta, adam, sgd
from keras import backend as K
from matplotlib.pyplot import imshow   #顯示圖片的function
import numpy as np    #運算工具

# loading train and test sets and reshape
(x_train, y_train), (x_test, y_test) = mnist.load_data()   #讀取data
x_train = x_train.reshape(60000, 784)    #x=data, y=label, reshape (28x28) -> (784,)
x_test = x_test.reshape(10000, 784)    #60000 training data, 10000 test data
x_train = x_train.astype('float32')    #由於像數值為0~255的整數, 把它轉為float作計算
x_test = x_test.astype('float32')
x_train /= 255    #把0~255/255即是把數字轉為0~1, 意義是運算的數值變小
x_test /= 255

num_classes = 10
# convert class vectors to binary class matrices
y_train = keras.utils.to_categorical(y_train, num_classes)    #把數字由0~9轉為one-hot編碼
y_test = keras.utils.to_categorical(y_test, num_classes)


model = models.Sequential()
model.add(layers.Dense(1024, input_shape=(784,)))    #first layer need to input data shape
model.add(layers.Activation("relu"))    #neural network一般使用relu
model.add(layers.Dense(128))     #dense(128) include 128 parameters in a layer
model.add(layers.Activation("relu"))
model.add(layers.Dense(10))    #最後一層等於分類數目
model.add(layers.Activation("softmax"))    #最後一層使用softmax 
model.summary()

model.compile(loss='categorical_crossentropy',   #多分類使用categorical_crossentropy
              optimizer='rmsprop',    #optimizer一般有adam, adadelta, rmsprop, sgd等等  
              metrics=['accuracy'])    

batch_size = 128    #batch_size越大每次訓練的數據越多, 注意記憶體可能會不夠用  
num_classes = 10    
epochs = 10    #重覆訓練10輪
history = model.fit(x_train, y_train,    #model.fit才是正式訓練模型, 上面只是設定model
                    batch_size=batch_size,
                    epochs=epochs,
                    verbose=1,      #verbose=1表示顯示進度條, 共有0,1,2三種模式
                    validation_data=(x_test, y_test))   

1
2
3
4
5
6
7
8
(x_train, y_train), (x_test, y_test) = mnist.load_data()    #下載mnist數據集
imshow(x_train[0], cmap='gray')    #cv2中用來顯示灰色圖片的語法  
print("x_train.shape:", x_train.shape)    #所有data的shape  
print("x_train[0].shape:", x_train[0].shape)    #第一個data的shape 
print("y_train.shape:", y_train.shape)    #training data的label的shape 
print("y_test.shape:", y_test.shape)    #test data的label的shape
print("y_train:", y_train)    #training data的label
print("y_test:", y_test)    #testing data的label

1
2
3
4
print("y_train[0]:", y_train[0])    #原始的label  
y_train = keras.utils.to_categorical(y_train, num_classes)    #one-hot後的label  
y_test = keras.utils.to_categorical(y_test, num_classes)    #one-hot後的label  
print("y_train[0]:", y_train[0])