{"id":"arxiv/2609.17845","url":"https://arxiv.org/abs/2609.17845","tags":["stat.ML","cs.LG","math.ST"],"score":0,"title":"Sharp margin-based generalization bounds for realizable SVM","domain":"stat.ML","source":"arxiv","authors":["Steve Hanneke","Aryeh Kontorovich"],"reviews":0,"subject":"stat.ML","summary":"Let the exact homogeneous hard-margin support vector machine be trained on \\(m\\) independent observations from a Borel probability law on a real Hilbert space. We prove that, with score zero counted as an error, there is a universal numerical constant \\(C\\) such that \\[ \\Pp\\left( γ_m>0,\\quad \\Risk(u_m)> \\frac{C}{m} \\left( K_m+\\log\\frac1δ \\right) \\right) \\le δ. \\] Here \\(γ_m\\) is the empirical homogeneous margin, \\(u_m\\) is the exact minimum-norm unit-margin separator, \\(r_m\\) is the largest training radius, and \\(K_m:=r_m^2\\norm{u_m}^2=r_m^2/γ_m^2\\) on \\(\\{γ_m>0\\}\\). The proof is driven by a…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-09-15","reactions":{}}