A convolutional neural network-based classifier for the MNIST handwritten digit dataset with Tensorflow
the model we're going to work out in this example is:
Model_1 link
Input -> CONV (-> ReLU -> Pool) -> FC -> ReLU -> FC -> Softmax -> Loss
Accuracy = 99.07 %
Model_2 link
Input -> CONV (-> RELU -> Pool) -> CONV (-> RELU -> Pool) -> FC -> ReLU -> Dropout -> FC -> Softmax -> Loss Accuracy = 99.25 %