PnP-CF: A Consensus Flow-Matching Solver for General Inverse Problems

Key points

• Plug-and-Play (PnP) methods using continuous-time deep generative models as learned data distribution priors enable synthesizing high-quality solutions of model-based inverse problems.1
• The quality of the solutions depends critically on the interplay and consensus between the inverse problem solver and the generative model sampler.1
• PnP-CF combines a flow-matching denoiser with (i) ADMM-like dual variable updates and (ii) a time-adaptive pullback onto the intermediate flow probability path that combines a stochastic and a deterministic component.1

Sources (1)

• [1]PnP-CF: A Consensus Flow-Matching Solver for General Inverse Problems
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 08:23 AM

  We revisit a recently introduced flow-based PnP framework, which we refer to as PnP-CF (Consensus Flow), and extend its application to nonlinear inverse problems, including those with learned neural-network forward operators.

  Plug-and-Play (PnP) methods using continuous-time deep generative models as learned data distribution priors enable synthesizing high-quality solutions of model-based inverse problems.

Extractive summary: sentences quoted from the sources.