We started this chapter by learning what autoencoders are and how autoencoders are used to reconstruct their own input. We explored convolutional autoencoders, where instead of using feedforward networks, we used convolutional and deconvolutional layers for encoding and decoding, respectively. Following this, we learned about sparse which activate only certain neurons. Then, we learned about another type of regularizing autoencoder, called a contractive autoencoder, and at the end of the chapter, we learned about VAE which is a generative autoencoder model.

In the next chapter, we will learn about how to learn from a less data points using few-shot learning algorithms.