We introduce an invertible Wasserstein slicing map that embeds multivariate distributions into a Hilbert space, enabling functional data analysis with theoretical guarantees and application to real-world data.

Training-free approach for image generation with diffusion models conditioned on the color distribution of a reference image.