Multi-Objective Causal Bandits:策略层面学习与最小干预空间新界定
Multi-Objective Causal Bandits: Minimal Intervention Space and Policy-Level Learning
搞多目标决策或 bandit 研究的可以看看。它证明了被随机策略占优这个此前被忽略的现象,还把干预空间剪枝做到了图上的充要刻画,配套的 UCB 算法有对数遗憾证明,做公平性或多约束场景的会用得上。
一篇 arXiv 论文研究了带多个奖励节点的多目标因果 bandit(multi-objective causal bandits),动作对应因果图上变量子集的干预,性能用 Pareto 最优性评估。
作者指出仅靠成对 Pareto 占优比较不够:一个干预可能不被任何单个干预占优,却会被其他干预的随机化策略(凸组合)占优。论文给出可能 Pareto 最优最小干预集(minimal intervention sets)的充要图刻画,纠正了先前会纳入永不 Pareto 最优干预集的公式化,从图结构即可得到最小且可靠的候选干预族。
在此基础上,作者设计了 UCB 风格的在线算法,用与其他动作凸组合的占优测试淘汰动作,证明了依赖间隔的对数级 Pareto 遗憾;还针对随机化策略才能同时实现可行与最优的约束场景,提出带次线性遗憾与约束违反保证的约束 bandit 算法。
Multi-Objective Causal Bandits: Minimal Intervention Space and Policy-Level Learning
Many decision-making problems involve multiple objectives, where actions affect several performance metrics, fairness criteria, or safety constraints. We study multi-objective causal bandits with multiple reward nodes, where actions correspond to interventions on subsets of variables in a causal graph and performance is evaluated through Pareto optimality. Existing approaches to multi-objective bandits often rely on pairwise Pareto dominance or optimize a fixed scalarization. However, pairwise comparisons alone can fail to identify actions dominated by randomized policies, whose expected reward vectors lie in the convex hull of action-level rewards. Thus, an intervention may be suboptimal even if no single intervention dominates it, because it can be dominated by a randomized policy over other interventions. We formalize this policy-level phenomenon and show that it has consequences for both causal action-space reduction and online learning. First, we provide a necessary-and-sufficient graphical characterization of possibly Pareto- optimal minimal intervention sets. Our characterization yields a minimal, sound, and complete candidate intervention family from the graph alone, correcting prior formulations that may include intervention sets that are never Pareto-optimal under any compatible structural causal model. Second, over this reduced intervention space, we design an online UCB-style algorithm that eliminates actions using dominance tests against convex combinations of other actions, and prove logarithmic gap-dependent Pareto regret. Finally, we study constrained multi-objective causal bandits, where feasibility and optimality may be realized only with randomized policies, and develop an online UCB-style constrained bandit algorithm with sublinear regret and constraint-violation guarantees.
本条情报归属 AITOP「论文」核心追踪栏目。在跨区域产业观察坐标中,该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:因果推断、多目标优化、bandit、Pareto最优,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。
信源核验说明:本文原始线索采编自「arXiv cs.LG」(@Muhammad Qasim Elahi, Mahsa Ghasemi, Murat Kocaoglu),经 AITOP 多维评分管线结构化提炼。