Machine Learning in Elixir

Learning Without Supervision in Practice

So far in this chapter, you’ve learned a bit about deep unsupervised learning mostly from the perspective of deep generative modeling. You progressed from compressing data with autoencoders to generating data with VAEs and finally to generating data with GANs. You were able to train some models to generate decent-looking, low-resolution, handwritten digits—which you might think is exciting. But at the same time, you might still be wondering about the secrets of power models like DALL-E and why learning about GANs, VAEs, and Autoencoders is even useful. In this section, you’ll dive a bit deeper into modern unsupervised learning in practice.

Unsupervised Learning in Theory

Yann Lecun, one of the godfathers ...