Beginning Anomaly Detection Using

Figure 4-26.  Code to create the neural network

Figure 4-27.  Model graph shown in TensorBoard

Chapter 4   autoenCoders

---

144

 Convolutional  Autoencoders

Whenever your inputs are images, it makes sense to use convolutional neural networks 

(convnets or CNNs) as encoders and decoders. In practical settings, autoencoders 

applied to images are always convolutional autoencoders because they simply perform 

much better.

Let’s implement one. The encoder will consist in a stack of Conv2D and MaxPooling2D 

layers (max pooling is being used for spatial down-sampling), while the decoder will 

consist in a stack of Conv2D and UpSampling2D layers.

Figure 

4-28

 shows the basic code to import all necessary packages in a Jupyter 

notebook. Also note the versions of the various packages.

Figure 4-28.  Importing packages in a Juypter notebook

Chapter 4   autoenCoders

---

145

You will use the mnist images data set for this purpose. Mnist contains images for the 

digits 0 to 9 and is used for many different use cases. Figure 

4-29

 shows the code to load 

MNIST data.