{"id":"arxiv/2606.14954","url":"https://arxiv.org/abs/2606.14954","tags":["math.FA","cs.LG","math.OC","stat.ML"],"score":0,"title":"Representation Costs in Data Science: Foundations and the Quasi-Banach Spaces of Deep Neural Networks","domain":"math.FA","source":"arxiv","authors":["Greg Ongie","Rahul Parhi"],"reviews":0,"subject":"math.FA","summary":"We develop a general framework for analyzing representation costs induced by parameter-space regularizers in data-fitting methods. For an arbitrary parametric method, we define its representation cost and native function space, prove existence, and identify conditions under which parameter-space and function-space problems have equal infimal values and minimizers transfer between them. This framework yields representer theorems and recovers classical formulations---including kernel methods and RKHSs, wavelets and Besov spaces, and shallow neural networks and variation spaces---as special…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-06-12","reactions":{}}