Cover Image for Diffusion Model Paper Reading Group Debut #2 - Unified Perspective on Diffusion Models: Connecting VAE, VDM, and SGM
Cover Image for Diffusion Model Paper Reading Group Debut #2 - Unified Perspective on Diffusion Models: Connecting VAE, VDM, and SGM
Avatar for AI Scholars
Presented by
AI Scholars
28 Going

Diffusion Model Paper Reading Group Debut #2 - Unified Perspective on Diffusion Models: Connecting VAE, VDM, and SGM

Virtual
Registration
Welcome! To join the event, please register below.
About Event

🗓 Agenda

[10 min] Welcome & community check-in

[40 min] 📄 Paper Discussion:

– Explain what is, and talk through the evolution from ELBO & VAEs → Variational Diffusion Models → Score-Based Generative Models

– Key ideas from “Understanding Diffusion Models: A Unified Perspective” (Calvin Luo, 2023)

– How different diffusion frameworks connect: DDPM, SDE, Score-based models

[10 min] Community Q&A + discussion

[5 min] What’s next in the Diffusion Study Group


📚 Pre-Class Learning

Skim this paper to get the most out of our session:

🧾 Understanding Diffusion Models: A Unified Perspective (Calvin Luo, 2023)

Link: https://arxiv.org/abs/2208.11970


🧠 Why This Paper Matters

This paper gives one of the clearest overviews of the evolution and unification of generative diffusion models — ELBO-based models like VAEs, to Variational Diffusion Models (VDMs), and finally to Score-Based Generative Models (SGMs) — showing how they are connected under a unified mathematical perspective.

Ideal if you’ve been confused by the variety of approaches in GenAI — and want to build a deep foundation before tackling advanced models like Diffusion Transformers or Flow Matching.


About the Diffusion Model Bootcamp

This diffusion model paperless reading group is a part of a 5-month, peer-led learning journey inspired by MIT’s diffusion model & flow matching course — independently designed to suit the needs of engineers, PMs, and curious builders.

🔹 You don’t need an ML background — just an interest in GenAI and a few focused hours a week.

🔹 We help you understand the landscape of diffusion model architectures — from basic ODE math concepts to the most important models related to diffusion models such as VAEs, DDPM, Flow Matching, and Diffusion Transformers.

🔹 You’ll also train your own model, build a real GenAI app, and explore modern tools like ComfyUI.

📚 Up-to-date curriculum | 💬 Monthly paper readings

🤝 Discussion-style 12 ppl class & supportive | 🧪 Final project


📅 Next cohort begins October 2025

📣 Limited spots — join this paper session to see if it’s the right fit for you!

👉 Questions? DM the organizer!

Avatar for AI Scholars
Presented by
AI Scholars
28 Going