Diffusion Model from Scratch

TL;DR

This paper provides a comprehensive, step-by-step explanation of diffusion generative models, tracing their evolution from VAEs to DDPM, aimed at helping students understand the complex underlying concepts.

Contribution

It offers detailed mathematical derivations and an analytical approach to understanding diffusion models, making the topic accessible for learners.

Findings

Clarifies the evolution from VAEs to DDPM

Highlights core ideas and improvement strategies

Provides educational guidance for students

Abstract

Diffusion generative models are currently the most popular generative models. However, their underlying modeling process is quite complex, and starting directly with the seminal paper Denoising Diffusion Probability Model (DDPM) can be challenging. This paper aims to assist readers in building a foundational understanding of generative models by tracing the evolution from VAEs to DDPM through detailed mathematical derivations and a problem-oriented analytical approach. It also explores the core ideas and improvement strategies of current mainstream methodologies, providing guidance for undergraduate and graduate students interested in learning about diffusion models.