{"id":"arxiv/2611.15492","url":"https://arxiv.org/abs/2611.15492","tags":["math.OC"],"score":0,"title":"Probabilistically Exact Convexification of Mixed-Integer Quadratic Optimization with Indicators","domain":"math.OC","source":"arxiv","authors":["Salar Fattahi","Andres Gomez"],"reviews":0,"subject":"math.OC","summary":"We study the convexification of mixed-integer quadratic optimization with indicator variables. The problem remains NP-hard even when the quadratic matrix has bandwidth two, posing a fundamental obstacle to tractable exact convexification. To overcome this challenge in realistic settings, we introduce probabilistically exact (inner) convexification, a framework that relaxes the requirement of exactness for every objective while preserving exact optimality for statistically relevant ones. Our approach is motivated by inference applications, where the objective is determined by a data vector…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-11-11","reactions":{}}