Classifier-Free Diffusion Guidance
TLDR
This paper introduces classifier-free guidance, a method to improve conditional diffusion model sampling quality without needing a separate classifier by jointly training conditional and unconditional models.
Key contributions
- Proposes classifier-free guidance that eliminates the need for a separate image classifier in diffusion model guidance.
- Jointly trains conditional and unconditional diffusion models to combine their score estimates for improved sample quality and diversity.
- Achieves a trade-off between mode coverage and sample fidelity comparable to traditional classifier guidance methods.
Why it matters
This paper matters because it simplifies the guidance process in conditional diffusion models by removing the dependency on an external classifier, reducing complexity and training overhead while maintaining high-quality and diverse sample generation. This advancement broadens the applicability and efficiency of diffusion-based generative models in various tasks.
Original Abstract
Classifier guidance is a recently introduced method to trade off mode coverage and sample fidelity in conditional diffusion models post training, in the same spirit as low temperature sampling or truncation in other types of generative models. Classifier guidance combines the score estimate of a diffusion model with the gradient of an image classifier and thereby requires training an image classifier separate from the diffusion model. It also raises the question of whether guidance can be performed without a classifier. We show that guidance can be indeed performed by a pure generative model without such a classifier: in what we call classifier-free guidance, we jointly train a conditional and an unconditional diffusion model, and we combine the resulting conditional and unconditional score estimates to attain a trade-off between sample quality and diversity similar to that obtained using classifier guidance.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.