Abstract:Reinforcement learning (RL) has been widely applied in recommendation systems due to its potential in optimizing the long-term engagement of users. From the perspective of RL, recommendation can be formulated as a Markov decision process (MDP), where recommendation system (agent) can interact with users (environment) and acquire feedback (reward signals).However, it is impractical to conduct online interactions with the concern on user experience and implementation complexity, and we can only train RL recommenders with offline datasets containing limited reward signals and state transitions. Therefore, the data sparsity issue of reward signals and state transitions is very severe, while it has long been overlooked by existing RL <a class="link-external link-http" href="http://recommenders.Worse" rel="external noopener nofollow">this http URL</a> still, RL methods learn through the trial-and-error mode, but negative feedback cannot be obtained in implicit feedback recommendation tasks, which aggravates the overestimation problem of offline RL recommender. To address these challenges, we propose a novel RL recommender named model-enhanced contrastive reinforcement learning (MCRL). On the one hand, we learn a value function to estimate the long-term engagement of users, together with a conservative value learning mechanism to alleviate the overestimation <a class="link-external link-http" href="http://problem.On" rel="external noopener nofollow">this http URL</a> the other hand, we construct some positive and negative state-action pairs to model the reward function and state transition function with contrastive learning to exploit the internal structure information of MDP. Experiments demonstrate that the proposed method significantly outperforms existing offline RL and self-supervised RL methods with different representative backbone networks on two real-world datasets.

Click is Not Equal to Purchase: Multi-task Reinforcement Learning for Multi-behavior Recommendation

Multi-Task Recommendations with Reinforcement Learning

Learning Recommender Systems from Multi-Behavior Data

Learning to Recommend with Multiple Cascading Behaviors

Neural Multi-task Recommendation from Multi-behavior Data.

MaHRL: Multi-goals Abstraction based Deep Hierarchical Reinforcement Learning for Recommendations

Model-enhanced Contrastive Reinforcement Learning for Sequential Recommendation

Behavior Pattern Mining-based Multi-Behavior Recommendation

Multi-view Multi-behavior Contrastive Learning in Recommendation

AdaRec: Adaptive Sequential Recommendation for Reinforcing Long-term User Engagement

Dual Contrastive Learning for Multi-Behavior Recommendation

Multi-Task Fusion via Reinforcement Learning for Long-Term User Satisfaction in Recommender Systems

Intrinsically Motivated Reinforcement Learning Based Recommendation with Counterfactual Data Augmentation

Modelling Temporal Dynamics and Repeated Behaviors for Recommendation

Knowledge Enhancement for Contrastive Multi-Behavior Recommendation

Multi-behavior multi-task recommendation algorithm integrating self-supervised learning enhancement

Multi-objective Meta-return Reinforcement Learning for Sequential Recommendation

Whole-Chain Recommendations

An Enhanced-State Reinforcement Learning Algorithm for Multi-Task Fusion in Large-Scale Recommender Systems

A Multi-Behavior Recommendation Method Exploring the Preference Differences among Various Behaviors.

Multi-Relational Contrastive Learning for Recommendation