Use 28x28 pixel images to recognize digit 0-9
Credit: Deep Learning with Python by Jason Brownlee
flattern and normalize the input
MLP with two dense layers
categorical cross entropy as the loss function, adam as the optimizer, and accuracy as the metrics.
10 epochs with batch size = 200
Baseline Error: 1.67%
reshape and normalize the input
simple CNN with 1 convolutional layer, 1 pooling, 1 dropout, 1 flattern, 2 dense layers
Much slower than MLP
categorical cross entropy as the loss function, adam as the optimizer, and accuracy as the metrics.
10 epochs with batch size = 200
CNN Error: 1.09% reduced error rate
larger CNN with 2 convolutional+pooling layers, 1 dropout, 1 flattern, 3 dense layers
Much slower than MLP
categorical cross entropy as the loss function, adam as the optimizer, and accuracy as the metrics.
10 epochs with batch size = 200
Large CNN Error: 0.77%, improved!