Deep neural network for recognizing handwritten digits. Uses the MNIST database of digits for training and validation.
This model has a 3 layer model with 16, 16 hidden units. Achieves accuracy above 96%
This model uses a convolutional neural network with 1 convolutional layers of kernel size 7x7, followed by Adam optimization and Batch Normalization, which is then fed into a MaxPool layer, this is repeated twice, followed by two fully connected layers. This model achieves greater than 99.2% accuracy.