A TensorFlow model of Convolutional neural network for digit recognition using MNIST data set.
Accuracy = 97.36% on test set with 1000 epoch on mini batch of 50 using AdamOptimizer.
Structure of Network
- Input data
- Convolution and max pooling.
- Convolution and max pooling.
- Fully connected layer.
- Processing - DropOut.
- Readout layer - Fully Connected.
- Output - Classified digits.