跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.21880 · v1

Cancellation-Free Constrained SquareCB for Safe Controller Selection

首次提交 2026-11-17 04:47
版本更新 2026-11-17 04:47
雷达观测 2026-11-18 19:44
arXiv ↗PDF ↗

摘要

We study constrained contextual bandits, in which a learner chooses one of KK actions in each round to minimize loss while keeping the expected cost of its play below a budget. Existing reductions to online regression weight the cost by a Lagrange multiplier learned from past constraint residuals, so slack in some rounds can offset violations in others. We instead run SquareCB proposed in \citep{foster2020beyond} with a rectified weight on the cost: it is zero when the estimated cost is within the budget, and otherwise just large enough to bring it back to the budget. Under realizability and Slater's condition with margin δδ, the algorithm attains regret O(KT RegSq(T)/δ)O(\sqrt{KT\,\mathrm{Reg}_{\mathsf{Sq}}(T)}/δ) and cancellation-free violation ∑t(Vcpt−b)+=O(KT RegSq(T))\sum_t(V_c^{p_t}-b)^+=O(\sqrt{KT\,\mathrm{Reg}_{\mathsf{Sq}}(T)}), where RegSq\mathrm{Reg}_{\mathsf{Sq}} is the error of the regression oracle. The same bounds hold when the budget must be met in every context, potentially adversarially chosen contexts. In controller selection for a safe-control-gym quadrotor and a BOPTEST building, the violation is 6 to 9 times lower than that of primal-dual and virtual-queue baselines, with a lower or comparable regret.

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-17 04:47 · 最新观测版本

Cancellation-Free Constrained SquareCB for Safe Controller Selection

We study constrained contextual bandits, in which a learner chooses one of KK actions in each round to minimize loss while keeping the expected cost of its play below a budget. Existing reductions to online regression weight the cost by a Lagrange multiplier learned from past constraint residuals, so slack in some rounds can offset violations in others. We instead run SquareCB proposed in \citep{foster2020beyond} with a rectified weight on the cost: it is zero when the estimated cost is within the budget, and otherwise just large enough to bring it back to the budget. Under realizability and Slater's condition with margin δδ, the algorithm attains regret O(KT RegSq(T)/δ)O(\sqrt{KT\,\mathrm{Reg}_{\mathsf{Sq}}(T)}/δ) and cancellation-free violation ∑t(Vcpt−b)+=O(KT RegSq(T))\sum_t(V_c^{p_t}-b)^+=O(\sqrt{KT\,\mathrm{Reg}_{\mathsf{Sq}}(T)}), where RegSq\mathrm{Reg}_{\mathsf{Sq}} is the error of the regression oracle. The same bounds hold when the budget must be met in every context, potentially adversarially chosen contexts. In controller selection for a safe-control-gym quadrotor and a BOPTEST building, the violation is 6 to 9 times lower than that of primal-dual and virtual-queue baselines, with a lower or comparable regret.

查看此版本解读 →查看此版本原文 ↗