MNIST 手写识别 + keras 框架

我是一个机器学习和神经网络学习的菜鸟，在网上找了各种资料进行学习，希望能快速提高自己的能力。
下面贴一份用 keras 模块解决 MNIST 手写识别的代码，供研究，也防止自己弄丢了

python 版本 ： 3.6.0
keras 版本： 2.1.3

# 代码来源
# http://www.cnblogs.com/Mu001999/p/6221096.html
import numpy as np
import gzip
import struct
import keras as ks
import logging
from keras.layers import Dense, Activation, Flatten, Convolution2D
from keras.optimizers import SGD, Adadelta,Adagrad
from keras.utils import np_utils

def read_data(label_url,image_url):
    with gzip.open(label_url) as flbl:
        magic, num = struct.unpack(">II",flbl.read(8))
        label = np.fromstring(flbl.read(),dtype=np.int8)
    with gzip.open(image_url,'rb') as fimg:
        magic, num, rows, cols = struct.unpack(">IIII",fimg.read(16))
        image = np.fromstring(fimg.read(),dtype=np.uint8).reshape(len(label),rows,cols)
    return (label, image)

# 训练集的数据
(train_lbl, train_img) = read_data('train-labels-idx1-ubyte.gz','train-images-idx3-ubyte.gz')
print("train_img.shape : " )
print( train_img.shape)
print("train_lbl.shape : " )
print( train_lbl.shape)

# 验证集的数据
(val_lbl, val_img) = read_data('t10k-labels-idx1-ubyte.gz','t10k-images-idx3-ubyte.gz')
print("val_img.shape : " )
print (val_img.shape)
print("val_lbl.shape : " )
print (val_lbl.shape)

print("读取数据完成")

# 将二维数据转换成一维数据 （原本是28 * 28 的， 转换成 784 的向量）
def to4d(img):
    return img.reshape(img.shape[0],784).astype(np.float32)/255

train_img = to4d(train_img)
val_img = to4d(val_img)
# label 向量转换成 0-1 型矩阵向量
#  如 a = [1,2,3]
#     b = np_utils.to_categorical(a,num_classes = 3)
#  则 b = [[0,1,0,0],
#          [0,0,1,0],
#          [0,0,0,1]] 
train_LBL = np_utils.to_categorical(train_lbl,num_classes=10)
val_LBL = np_utils.to_categorical(val_lbl,num_classes=10)

model = ks.models.Sequential()
###################第一套可用设置########################
model.add(Dense(128,input_dim=784))
model.add(Activation('relu'))
model.add(Dense(64))
model.add(Activation('relu'))
model.add(Dense(10))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy',optimizer='adadelta',metrics=['accuracy'])
model.fit(x=train_img,y=train_LBL,batch_size=100,nb_epoch=10,verbose=1,validation_data=(val_img,val_LBL))
###################第一套可用设置########################

执行结果 ：

Using TensorFlow backend.
train_img.shape :
(60000, 28, 28)
train_lbl.shape :
(60000,)
val_img.shape :
(10000, 28, 28)
val_lbl.shape :
(10000,)
读取数据完成
C:\Users\admin\AppData\Local\Programs\Python\Python36\lib\site-packages\keras\models.py:944: UserWarning: The `nb_epoch` argument in `fit` has been renamed `epochs`.
  warnings.warn('The `nb_epoch` argument in `fit` '
Train on 60000 samples, validate on 10000 samples
Epoch 1/10
2018-01-31 10:08:59.666247: I C:\tf_jenkins\home\workspace\rel-win\M\windows\PY\36\tensorflow\core\platform\cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2
60000/60000 [==============================] - 3s 52us/step - loss: 0.3090 - acc: 0.9094 - val_loss: 0.1502 - val_acc: 0.9536
Epoch 2/10
60000/60000 [==============================] - 3s 43us/step - loss: 0.1311 - acc: 0.9609 - val_loss: 0.1048 - val_acc: 0.9666
Epoch 3/10
60000/60000 [==============================] - 3s 45us/step - loss: 0.0914 - acc: 0.9722 - val_loss: 0.0918 - val_acc: 0.9706
Epoch 4/10
60000/60000 [==============================] - 3s 45us/step - loss: 0.0713 - acc: 0.9781 - val_loss: 0.0842 - val_acc: 0.9734
Epoch 5/10
60000/60000 [==============================] - 3s 45us/step - loss: 0.0567 - acc: 0.9825 - val_loss: 0.0855 - val_acc: 0.9731
Epoch 6/10
60000/60000 [==============================] - 3s 46us/step - loss: 0.0464 - acc: 0.9858 - val_loss: 0.0856 - val_acc: 0.9730
Epoch 7/10
60000/60000 [==============================] - 3s 45us/step - loss: 0.0389 - acc: 0.9886 - val_loss: 0.0734 - val_acc: 0.9770
Epoch 8/10
60000/60000 [==============================] - 3s 45us/step - loss: 0.0326 - acc: 0.9905 - val_loss: 0.0756 - val_acc: 0.9771
Epoch 9/10
60000/60000 [==============================] - 3s 45us/step - loss: 0.0271 - acc: 0.9919 - val_loss: 0.0705 - val_acc: 0.9782
Epoch 10/10
60000/60000 [==============================] - 3s 46us/step - loss: 0.0223 - acc: 0.9935 - val_loss: 0.0713 - val_acc: 0.9787