Summary

Autoencoders are a great tool for unsupervised learning from data. They are often used for dimensionality reduction so that data can be represented by the lesser number of features. In this chapter, you learned about various types of autoencoders. We practiced building the three types of autoencoders using TensorFlow and Keras: stacked autoencoders, denoising autoencoders, and variational autoencoders. We used the MNIST dataset as an example.

In the last chapters, you have learned how to build various kinds of machine learning and deep learning models with TensorFlow and Keras, such as regression, classification, MLP, CNN, RNN, and autoencoders. In the next chapter, you will learn about advanced features of TensorFlow and Keras that allow us to take the models to production.