Some trics for kaggle mnist data recognition competition
- python3
- tensorflow
- keras
- marplotlib
- scipy
Input: (60000, 784, 1) (60000 images shape = 28x28)
Hidden: Iteratively changing hidden layer size
Output: One hot vector with 10 labels (0..9)
Rotate digit to vertical and stretch
For human it makes digits less recognisible, but for NN digits start to look more similar
!important: no improovement in case of usage CNN