Score Function Interpolation For Guidance Computation

* ⭐ 08/2022: [Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation (DreamBooth)](https://arxiv.org/abs/2208.12242)

Unique: Uses direct linear interpolation in score function space (s_guided = s_cond + w*(s_cond - s_uncond)) rather than gradient-based guidance or classifier-based methods, enabling simple, efficient computation without external models or gradient computation

vs others: Computationally simpler and faster than classifier guidance (no gradient computation through external classifier) and more direct than adversarial guidance methods, but assumes score function compatibility and requires careful scale tuning