方法对比

	扩散模型迁移学习	自监督扩散模型
领域	深度学习	深度学习
方法族	Machine learning	Machine learning
起源年份	2020–2023	2020–2022
提出者	Ho et al. (DDPM); transfer application popularized by Rombach et al. (Stable Diffusion) and Ruiz et al. (DreamBooth), 2020–2023	Ho, J. et al.; extended by Chen, T. et al. and subsequent self-supervised diffusion works
类型	Generative model with transfer learning	Generative model with self-supervised representation objective
开创性文献	Ho, J., Jain, A., & Abbeel, P. (2020). Denoising Diffusion Probabilistic Models. Advances in Neural Information Processing Systems (NeurIPS), 33, 6840–6851.	Ho, J., Jain, A., & Abbeel, P. (2020). Denoising Diffusion Probabilistic Models. Advances in Neural Information Processing Systems (NeurIPS), 33, 6840–6851.
别名	diffusion model fine-tuning, pre-trained diffusion transfer, TL-DM, domain-adapted diffusion model	SSDM, self-supervised score-based model, diffusion-based self-supervised learning, denoising diffusion with self-supervised pretraining
相关	5	2
摘要	Transfer Learning with Diffusion Models adapts a large pre-trained diffusion model — such as Stable Diffusion or DALL-E 2 — to a new target domain or task by continuing training on a smaller domain-specific dataset. Rather than learning the full generative process from scratch, practitioners leverage knowledge already encoded in millions of training steps to achieve high-quality domain-adapted generation with modest data and compute.	A self-supervised diffusion model couples the iterative noise-and-denoise generative process of denoising diffusion probabilistic models with a self-supervised representation learning objective — such as contrastive or masked prediction loss — so that the model simultaneously learns to generate realistic data and to produce semantically meaningful representations without any labeled examples.