{"id":"arxiv/2610.09731","url":"https://arxiv.org/abs/2610.09731","tags":["math.ST","cs.LG","stat.ML"],"score":0,"title":"Unbounded Characteristic and Universal Kernels","domain":"math.ST","source":"arxiv","authors":["Jose Cribeiro-Ramallo","Florian Kalinke","Zoltán Szabó"],"reviews":0,"subject":"math.ST","summary":"Kernel methods are among the most powerful tools in machine learning and statistics, with a large number of successful applications. Their immense success stems from the flexible function class associated to each kernel---its reproducing kernel Hilbert space (RKHS)---which facilitates statistical analysis, as well as from their computational tractability and applicability to many domains. Multiple notions (such as characteristic, $L_p$-universal, and integrally strictly positive definite) capture the expressivity of kernels and their RKHSs and play a key role in understanding the statistical…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-07","reactions":{}}