Diffusion models are a type of generative model that have been used in several popular deep learning models, such as DALL-E 2. The primary function of diffusion models is to map training data to a latent space using a Markov chain. This process gradually adds noise to the data, resulting in an asymptotically transformed image that is Gaussian in nature. The ultimate goal of this process is to learn its reverse, which enables us to generate new data by traversing backwards.
Diffusion models have a wide range of applications, including text simplification, question generation, text-to-image generation, paraphrasing, and more. The purpose of this project is to explore some of these applications and potentially achieve further results