{"id":"arxiv/2608.27908","url":"https://arxiv.org/abs/2608.27908","tags":["math.PR","math.CO"],"score":0,"title":"A Two-regime Khintchine Inequality and an Improved Bound on the Degree-1 Fourier Weight for Linear Threshold Functions","domain":"math.PR","source":"arxiv","authors":["Xuan Fang","Tianyu Wang"],"reviews":0,"subject":"math.PR","summary":"The Khintchine inequality provides a lower bound on the expected absolute value of a weighted sum of independent Rademacher random variables. In the classical setting, when the weight vector has unit norm, this lower bound is a constant, with equality attained only for a simple family of extremal configurations. A refined version due to De, Diakonikolas, and Servedio (2013) -- referred to as the \\emph{linear Khintchine inequality} -- strengthens this by establishing a lower bound that depends linearly on the distance of the weight vector from the extremal set. In this paper, we present a…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-08-28","reactions":{}}