Independent Researcher, Independent
1 paper at NeurIPS 2025
Self-diffusion solves inverse problems without the need of pretrained generative models via a self-contained iterative process that alternates between noising and denoising steps to progressively refine its estimate of the solution.
