Understanding Diffusion: From SDE to Flow Matching
青稞AI · wechat · 2026-08-16
This in-depth article connects the dots between SDE, reverse SDE, probability flow ODE, and Flow Matching in the context of Diffusion models. Starting from first principles, it explores the essence of generation—transporting a noise distribution to a real data distribution—and explains why intermediate paths are necessary. The author derives the forward noising formula via linear SDE and uses time reversal theory to introduce reverse generation, naturally leading to the concept of Flow Matching.
More from Research
- GRPO Struggles with Complex Agents; Suggests Mixing Off-Policy RL — nagpalchirag · 2026-08-16
- Paper: Evidence from Nationwide Generative AI Rollout in Pakistan's Courts — soumitrashukla9 · 2026-08-16
- Starfield Fauna dataset released with 20k images — eccLykta · 2026-08-16
- Paper: miRNAs modulate bioelectrical regionalization in multicellular aggregates — drmichaellevin · 2026-08-16
- Intern-S2-Mobius Decouples Knowledge and Reasoning for 4x Speed — jiqizhixin · 2026-08-16
- Position Paper: Teaching as the Grand Challenge for Theory of Mind in AI — atilimgunes · 2026-08-16