Beginning Anomaly Detection Using

causal convolution on the input as part of the first 

convolutional layer. The outputs of the first convolutional layer, which you will call 

conv_1, are now the inputs of the first 

max pooling layer, which you will call pool_1.

Recall from Chapter 

3

 that the pooling layer emphasizes the maximum value in 

the areas it passes through, effectively generalizing the inputs by choosing the heaviest 

values. From here, you have another set of causal convolutions and max pooling with 

layers conv_2 and pool_2. Note the progressive reduction in size of the data as it passes 

through the encoding stage, a feature characteristic to autoencoders. Finally, you have a 

dense layer in the middle of the two stages, representing the final, encoded output of the 

encoding stage as well as the encoded input of the decoding stage.

Input