December 2018
Intermediate to advanced
764 pages
18h 18m
English
We keep the layers, filters, and their sizes the same as in the MNIST examples earlier, with one new addition of regularization layer. Since this data set is complex as compared to the MNIST, we add additional dropout layers for the purpose of regularization:
tf.nn.dropout(layer1_pool, keep_prob)
The complete code for LeNet model for CIFAR10 data is provided in the notebook ch-09b_CNN_CIFAR10_TF_and_Keras.
On running the model we get the following output:
Epoch: 0000 loss = 2.115784 Epoch: 0001 loss = 1.620117 Epoch: 0002 loss = 1.417657 Epoch: ...
Read now
Unlock full access