Generative Modeling with Diffusion
This is an incremental overview paper for researchers interested in generative modeling techniques.
The paper provides an overview of diffusion models for generating new samples, formally defining the noising and denoising processes and presenting training algorithms, with a potential application in improving classifier performance on imbalanced data.
We provide an overview of the diffusion model as a method to generate new samples. Generative models have been recently adopted for tasks such as art generation (Stable Diffusion, Dall-E) and text generation (ChatGPT). Diffusion models in particular apply noise to sample data and then "reverse" this noising process to generate new samples. We will formally define these noising and denoising processes, then present algorithms to train and generate with a diffusion model. Afterward, we will explore a potential application of diffusion models in improving classifier performance on imbalanced data.