{"id":"arxiv/2605.08777","url":"https://arxiv.org/abs/2605.08777","tags":["stat.ML","cs.LG","math.PR"],"score":0,"title":"Measuring and Decomposing Mode Separation via the Canonical Diffusion","domain":"stat.ML","source":"arxiv","authors":["Shaul Tolkovsky","Ori Meidler","Or Zuk"],"reviews":0,"subject":"stat.ML","summary":"Mode separation, namely how sharply a distribution fragments into barrier-separated clusters, is a fundamental geometric property of densities, difficult to quantify in high dimensions. It is structurally distinct from dispersion, yet existing tools fall short: differential entropy rises with spread regardless of fragmentation, PCA orders directions by variance regardless of barriers, and mutual information requires a mixture decomposition one usually does not have. We measure mode separation through a single stochastic process intrinsic to the density: a unique reversible diffusion with $f$…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-05-09","reactions":{}}