标题： 基于嵌入空间用户行为建模的排名策略的离策评估 显示英文标题

标题： Off-Policy Evaluation of Ranking Policies via Embedding-Space User Behavior Modeling

摘要： 在具有大规模排名动作空间的排名设置中的离线策略评估（OPE），由于唯一动作数量和排名长度的增加而变得至关重要，这使得仅使用先前版本的日志式乐队数据即可评估新的推荐器策略。 为了解决与现有估计量相关的高方差问题，我们引入了两个新假设：排名对排名没有直接影响以及用户行为模型在排名嵌入空间上。 然后，我们提出了广义边缘化逆倾向评分（GMIPS）估计器，与现有的相比具有统计上可取的特性。 最后，我们证明了GMIPS实现了最低的均方误差。 值得注意的是，在GMIPS变体中，边缘化奖励交互IPS（MRIPS）基于排名嵌入的级联行为假设，结合了双重边缘化的重要性权重。 即使排名动作空间增加且上述假设可能不成立，我们的实验表明，MRIPS有效地平衡了偏差和方差之间的权衡。 显示英文摘要

摘要： Off-policy evaluation (OPE) in ranking settings with large ranking action spaces, which stems from an increase in both the number of unique actions and length of the ranking, is essential for assessing new recommender policies using only logged bandit data from previous versions. To address the high variance issues associated with existing estimators, we introduce two new assumptions: no direct effect on rankings and user behavior model on ranking embedding spaces. We then propose the generalized marginalized inverse propensity score (GMIPS) estimator with statistically desirable properties compared to existing ones. Finally, we demonstrate that the GMIPS achieves the lowest MSE. Notably, among GMIPS variants, the marginalized reward interaction IPS (MRIPS) incorporates a doubly marginalized importance weight based on a cascade behavior assumption on ranking embeddings. MRIPS effectively balances the trade-off between bias and variance, even as the ranking action spaces increase and the above assumptions may not hold, as evidenced by our experiments.