| question | training accuracy | testing accuracy |
|---|---|---|
| Q1 | 97.52 | 88.20 |
| Q2 | 81.75 | 81.70 |
| Q3 | 98.02 | 86.10 |
| Q4 | 90.77 | 89.00 |
| Q5 | 99.68 | 99.19 |
使用LeNet网络架构 (reference：https://github.com/zergtant/pytorch-handbook/blob/master/chapter3/3.2-mnist.ipynb） （28，28） (convolution)-> (6, 24 , 24) (maxpooling) ->(6, 12, 12) (convolution) ->(16, 8, 8) (maxpooling) ->(16, 4, 4) (reshape) ->(256) (fuly connect) ->(120) (fuly connect) ->(84) (fuly connect) ->(10) (fuly connect)