<ul><li>Diffusion models struggle with meeting various constraints inherent in training data.</li><li>A new approach called neural approximate mirror maps (NAMMs) is proposed for handling general, possibly non-convex constraints in diffusion models.</li><li>NAMMs learn an approximate mirror map that transforms data into an unconstrained space, enabling reliable generation of valid synthetic data and solving constrained inverse problems.</li><li>Experimental results demonstrate the improvement in constraint satisfaction using NAMM-based mirror diffusion models (MDMs) compared to unconstrained diffusion models.</li></ul>

Neural Approximate Mirror Maps for Constrained Diffusion Models

Discover more