{"id":"arxiv/2610.39731","url":"https://arxiv.org/abs/2610.39731","tags":["cs.LG","cond-mat.dis-nn","math-ph"],"score":0,"title":"Going deep, reaching further: What can Boltzmann machines efficiently represent?","domain":"cs.LG","source":"arxiv","authors":["Tobias Reinhart","Gemma De les Coves"],"reviews":0,"subject":"cs.LG","summary":"While universal approximation theorems theoretically underpin the representation capabilities of machine learning models, they have two shortcomings: They do not focus on efficient representations and they tend to be insensitive to depth and other heuristically important properties. In this work we show that these two shortcomings are intertwined. We prove that machine learning models which are deeper or have higher-order interactions can efficiently represent a larger set of functions, i.e. have larger efficient reach. It follows that deep Boltzmann machines (DBMs) have maximal efficient…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-10-29","reactions":{}}