MNIST digits recognizer.
Results:
Reading data
60000/60000 Done
10000/10000 Done
Reading data done
Using dense model.
FlattenLayer Layer in shape: (*, 28, 28, 1) out shape: (*, 784)
Dense Layer in shape: (*, 784) out shape: (*, 512) total params: 401920
Activation Layer in shape: (*, 512) out shape: (*, 512)
DropoutLayer Layer in shape: (*, 512) out shape: (*, 512) rate: 0.300
Dense Layer in shape: (*, 512) out shape: (*, 512) total params: 262656
Activation Layer in shape: (*, 512) out shape: (*, 512)
DropoutLayer Layer in shape: (*, 512) out shape: (*, 512) rate: 0.300
Dense Layer in shape: (*, 512) out shape: (*, 10) total params: 5130
Activation Layer in shape: (*, 10) out shape: (*, 10)
total params: 669706
1 [================================] 1:06 train cost: 0.050571 test cost: 0.030399
2 [================================] 1:07 train cost: 0.024648 test cost: 0.026596
3 [================================] 1:06 train cost: 0.019629 test cost: 0.021218
4 [================================] 1:07 train cost: 0.015862 test cost: 0.018637
5 [================================] 1:06 train cost: 0.014168 test cost: 0.019536
6 [================================] 1:06 train cost: 0.012497 test cost: 0.018853
7 [================================] 1:06 train cost: 0.011439 test cost: 0.019084
8 [================================] 1:06 train cost: 0.010557 test cost: 0.019545
Valid: 9808 / 10000 [98.08%]