Introduction to Generative Models

Since we have reached the end of our journey, let us contemplate what we were expecting when we started. The goal was to be able to develop models that can do image- and sequence-related tasks efficiently and effectively. We now know that DL models can help us classify images and text. Sections III and IV of this book focus on the convolutional and sequential models that help us accomplish such tasks. We also learned to develop models that can accomplish slightly complex tasks like next character generation and encoding of an image. Let us now focus on more complex tasks and explore the fundamentals of generative models. Generative models not only help us carry out supervised and unsupervised learning tasks, studied so far, but also help us generate new data from a particular distribution. One of the glaring examples of generative models is ChatGPT, which has disrupted the field. It is based on transformers. This chapter introduces transformers. But before diving into transformers, let us have a basic idea of Hopfield Networks and Boltzmann Machines.