昨天的结果中，最好的效果是3层全连接：Train Accuracy: 1.0，Test Accuracy: 0.765
4层卷积网络
隐藏层：2层卷积层+1层全连接层，输出层：Softmax分类器
K = 4 #hindden1的卷积核数目
L = 8 #隐藏层2的卷积核数
N = 200 #全连接层的节点数
使用了学习率衰减：
max_lr = 0.003
min_lr = 0.0001
decay_speed = 2000.0
未使用Droupout。pkeep:1.0
iter = 800
All cost time: 1519.3290100097656
Train Accuracy: 1.0
Test Accuracy: 0.805
使用卷机网络耗时明显增加
核数目稍微增多
K = 5 #hindden1的卷积核数目
L = 10
pkeep:1.0
终于，loss曲线有了很大的改观
核数目减少
K = 3 #hindden1的卷积核数目
L = 5
All cost time: 1459.0826830863953
Train Accuracy: 1.0
Test Accuracy: 0.84
K = 2 #hindden1的卷积核数目
L = 2
All cost time: 1211.8775815963745
Train Accuracy: 1.0
Test Accuracy: 0.905
第一次上90%！
将全连接层的节点数减少
K = 2 #hindden1的卷积核数目
L = 2 #隐藏层2的卷积核数
N = 100
All cost time: 1124.0736384391785
Train Accuracy: 1.0
Test Accuracy: 0.76
N=500
All cost time: 1369.4422793388367
Train Accuracy: 1.0
Test Accuracy: 0.765
N=180
All cost time: 1144.1208844184875
Train Accuracy: 1.0
Test Accuracy: 0.95
终于，这个95%还算不错了
