Optimizing Long-term Value for Auction-Based Recommender Systems via On-Policy Reinforcement Learning
This addresses the issue of short-term optimization in online advertising platforms, potentially improving user engagement, but it is incremental as it builds on existing auction frameworks with reinforcement learning.
The paper tackled the problem of auction-based recommender systems focusing on immediate returns by using reinforcement learning to optimize for long-term user engagement metrics, and through an online A/B test on a system handling billions of daily impressions, it empirically outperformed the current production system.
Auction-based recommender systems are prevalent in online advertising platforms, but they are typically optimized to allocate recommendation slots based on immediate expected return metrics, neglecting the downstream effects of recommendations on user behavior. In this study, we employ reinforcement learning to optimize for long-term return metrics in an auction-based recommender system. Utilizing temporal difference learning, a fundamental reinforcement learning algorithm, we implement an one-step policy improvement approach that biases the system towards recommendations with higher long-term user engagement metrics. This optimizes value over long horizons while maintaining compatibility with the auction framework. Our approach is grounded in dynamic programming ideas which show that our method provably improves upon the existing auction-based base policy. Through an online A/B test conducted on an auction-based recommender system which handles billions of impressions and users daily, we empirically establish that our proposed method outperforms the current production system in terms of long-term user engagement metrics.