1.加入两个隐藏层并且加入DROPOUT功能
(1)输入层 784
(2)隐藏层1 1000 0.5 DropOut
(3)隐藏层2 1000 0.5 DropOut
(4)输出层 10
model = Sequential()
model.add(Dense(units=1000,input_dim=784,kernel_initializer=’normal’,activation=’relu’))
model.add(Dropout(0.5))
model.add(Dense(units=1000,kernel_initializer=’normal’,activation=’relu’))
model.add(Dropout(0.5))
model.add(Dense(units=10,kernel_initializer=’normal’,activation=’softmax’))
2.查看模型的摘要
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
dense_1 (Dense) (None, 1000) 785000
_________________________________________________________________
dropout_1 (Dropout) (None, 1000) 0
_________________________________________________________________
dense_2 (Dense) (None, 1000) 1001000
_________________________________________________________________
dropout_2 (Dropout) (None, 1000) 0
_________________________________________________________________
dense_3 (Dense) (None, 10) 10010
=================================================================
Total params: 1,796,010
Trainable params: 1,796,010
Non-trainable params: 0
_________________________________________________________________
None
3.查看训练过程的准确率
无论是训练还是验证,准确率都越来越高
在Epoch训练后期,虽然”acc训练的准确率”比”val_acc验证的准确率”高,但是”训练的准确率”与”验证的准确率”差距变小,这代表已经大致解决来过度拟合的问题
4.查看准确率
32/10000 […………………………] – ETA: 2s
384/10000 [>………………………..] – ETA: 1s
768/10000 [=>……………………….] – ETA: 1s
1184/10000 [==>………………………] – ETA: 1s
1600/10000 [===>……………………..] – ETA: 1s
1984/10000 [====>…………………….] – ETA: 1s
2304/10000 [=====>……………………] – ETA: 1s
2720/10000 [=======>………………….] – ETA: 0s
3136/10000 [========>…………………] – ETA: 0s
3552/10000 [=========>………………..] – ETA: 0s
3968/10000 [==========>……………….] – ETA: 0s
4320/10000 [===========>………………] – ETA: 0s
4672/10000 [=============>…………….] – ETA: 0s
4992/10000 [=============>…………….] – ETA: 0s
5344/10000 [===============>…………..] – ETA: 0s
5664/10000 [===============>…………..] – ETA: 0s
6016/10000 [=================>…………] – ETA: 0s
6368/10000 [==================>………..] – ETA: 0s
6752/10000 [===================>……….] – ETA: 0s
7136/10000 [====================>………] – ETA: 0s
7488/10000 [=====================>……..] – ETA: 0s
7872/10000 [======================>…….] – ETA: 0s
8224/10000 [=======================>……] – ETA: 0s
8608/10000 [========================>…..] – ETA: 0s
8992/10000 [=========================>….] – ETA: 0s
9376/10000 [===========================>..] – ETA: 0s
9792/10000 [============================>.] – ETA: 0s
10000/10000 [==============================] – 1s 140us/step
accuracy= 0.9803
准确率并没有显著的提升,如果还要进一步提升准确率,就必须使用卷积神经网络.