模型源自Yann LeCun(1998)的論文《Gradient-Based Learning Applied to Document Recognition》，用於MNIST資料集。模型輸入為32X32的灰度影象，第一層為6個5X5卷積核，不擴充套件邊界；第二層為2X2的最大值池化層，步進為2X2；第三層為16個5X5卷積核，不擴充套件邊界；第四層為2X2的最大值池化層，步進為2X2；第五層為展平層，並全連線120個節點；第六層為全連線層，84個節點；第七層為全連線softmax層，輸出結果。

        原論文中第二層池化層和第三層卷積層之間為是部分連線。本文中並未考慮，而是做成全連線，模型結構如下圖所示。

        模型採用keras的Sequential實現，源資料分為train和test兩個資料夾，每個資料夾下有十個子資料夾，分別方有各數字對應的灰度圖。實現程式碼如下：

import os
import cv2
from numpy import *
from keras.models import Sequential
from keras.layers import Dense
from keras.layers import Conv2D, MaxPooling2D, Flatten
from keras.optimizers import SGD
from keras.utils import np_utils
from keras.utils.vis_utils import plot_model

def loadData(path):
    data = []
    labels = []
    for i in range(10):
        dir = './'+path+'/'+str(i)
        listImg = os.listdir(dir)
        for img in listImg:
            data.append([cv2.imread(dir+'/'+img, 0)])
            labels.append(i)
        print path, i, 'is read'
    return data, labels


trainData, trainLabels = loadData('train')
testData, testLabels = loadData('test')
trainLabels = np_utils.to_categorical(trainLabels, 10)
testLabels = np_utils.to_categorical(testLabels, 10)


model = Sequential()
model.add(Conv2D(filters=6, kernel_size=(5,5), padding='valid', input_shape=(1,28,28), activation='tanh'))
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Conv2D(filters=16, kernel_size=(5,5), padding='valid', activation='tanh'))
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Flatten())
model.add(Dense(120, activation='tanh'))
model.add(Dense(84, activation='tanh'))
model.add(Dense(10, activation='softmax'))
sgd = SGD(lr=0.05, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(optimizer=sgd, loss='categorical_crossentropy', metrics=['accuracy'])
model.fit(trainData, trainLabels, batch_size=500, epochs=20, verbose=1, shuffle=True)

plot_model(model, to_file='model.png', show_shapes=True, show_layer_names=False)