Autoencoders are feed-forward, non-recurrent neural networks, which learn by unsupervised learning. They have an inherent capability to learn a compact representation of data. They are at the centre of deep belief networks and find applications in image reconstruction, clustering, machine translation, and much more. In this chapter, you will learn and implement different variants of autoencoders and eventually learn how to stack autoencoders. The chapter includes the following topics:

• Vanilla autoencoder
• Sparse autoencoder
• Denoising autoencoder
• Convolutional autoencoders
• Stacked autoencoders