{"id":"arxiv/2607.24007","url":"https://arxiv.org/abs/2607.24007","tags":["math.OC","cs.LG"],"score":0,"title":"Smooth Learning with Hard Constraints via Legendre-Regularized Policies","domain":"math.OC","source":"arxiv","authors":["Zikun Lin","Rui Chen","Yijie Wang"],"reviews":0,"subject":"math.OC","summary":"We revisit contextual optimization from the perspective of policy class design. A desirable policy class should be expressive enough to learn rich context-decision relationships, should enforce hard feasibility constraints rather than soft penalty terms, and should remain smooth enough for gradient-based training on downstream decision losses. Existing approaches usually emphasize only part of these requirements. We propose Legendre-regularized policies, which parameterize decisions as solutions of regularized optimization problems over the original feasible region. This construction yields…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-07-27","reactions":{}}