Summary

You have now learned how to build a VAE and about the advantages of using a VAE over a standard autoencoder. You have also learned about the effect of varying latent space dimensions on the network.

As an exercise, you should try training this model to work on the CIFAR-10 dataset and using convolutional layers instead of simple fully connected layers.

In the next chapter, we will look at what data pipelines are and why we use Pachyderm to build or manage them.