Abstract:As the last stage of a typical \textit{recommendation system}, \textit{collective recommendation} aims to give the final touches to the recommended items and their layout so as to optimize overall objectives such as diversity and whole-page relevance. In practice, however, the interaction dynamics among the recommended items, their visual appearances and meta-data such as specifications are often too complex to be captured by experts' heuristics or simple models. To address this issue, we propose a \textit{\underline{div}ersity-aware self-correcting sequential recommendation \underline{net}works} (\textit{DivNet}) that is able to estimate utility by capturing the complex interactions among sequential items and diversify recommendations simultaneously. Experiments on both offline and online settings demonstrate that \textit{DivNet} can achieve better results compared to baselines with or without collective recommendations.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to solve several key challenges faced by the collective recommendation module in recommendation systems, especially problems in optimizing the diversity of recommended items and the overall page relevance. Specifically: 1. **Complex Interactions**: - There are complex interactions among recommended items. These interactions depend not only on the order of the items but also on their visual appearance and metadata (such as specifications). Traditional rule - based or simple - model - based methods are difficult to capture these complex interactions. 2. **Differences between Training and Inference**: - During the training process, the model loss takes into account the interactions between items. However, during inference, items are scored one by one, or the context has changed, resulting in unreliable prediction performance. 3. **Lack of Diversity**: - Similar items usually get higher scores, leading to overly concentrated recommended results and a lack of diversity. This will affect the user experience and click - through rate. 4. **Sequential Exposure Bias**: - Only a small portion of possible item combinations are shown to users, making it unreliable to estimate the effects of unshown combinations. To address these problems, the authors propose a model named DivNet (Diversity - Aware Self - Correcting Sequential Recommendation Networks). DivNet solves the problems in the following ways: - **Capturing Complex Interactions**: Use the self - attention mechanism to capture complex interactions among items. - **Diversifying Recommendations**: Introduce a self - correcting process to ensure the diversity of the recommendation list. - **Reducing Exposure Bias**: Reduce the impact of sequential exposure bias through supervised regularization. Through these methods, DivNet can improve the diversity of recommended results while maintaining high relevance, thereby enhancing user satisfaction and click - through conversion rate. ### Summary The main objective of this paper is to solve the problems of complex interactions, differences between training and inference, lack of diversity, and sequential exposure bias faced by the collective recommendation module in recommendation systems by introducing the DivNet model, thereby optimizing the layout of recommended items and the overall page relevance.

DivNet: Diversity-Aware Self-Correcting Sequential Recommendation Networks

Co-training Disentangled Domain Adaptation Network for Leveraging Popularity Bias in Recommenders

Divrec: A Framework For Top-N Recommendation With Diversification In E-Commerce

Improving Accuracy and Diversity in Matching of Recommendation with Diversified Preference Network

Relieving Popularity Bias in Interactive Recommendation: A Diversity-Novelty-Aware Reinforcement Learning Approach

Dual-Process Graph Neural Network for Diversified Recommendation

Heterogeneous Graph Neural Network with Personalized and Adaptive Diversity for News Recommendation

Diversify or Not: Dynamic Diversification for Personalized Recommendation

Diversity-Promoting Deep Reinforcement Learning for Interactive Recommendation

End-to-end Learnable Diversity-aware News Recommendation

A Recommendation Approach Based on Heterogeneous Network and Dynamic Knowledge Graph

Multi-interest Diversification for End-to-end Sequential Recommendation

Recent Advances in Diversified Recommendation

Graph Exploration Matters: Improving both individual-level and system-level diversity in WeChat Feed Recommender

Personalized Re-ranking for Improving Diversity in Live Recommender Systems

Personalized Diversification for Neural Re-ranking in Recommendation.

Diversity-Aware Top-N Recommendation: A Deep Reinforcement Learning Way

DGCN: Diversified Recommendation with Graph Convolutional Networks

A Joint Optimization Approach for Personalized Recommendation Diversification.

Ada-Ranker: A Data Distribution Adaptive Ranking Paradigm for Sequential Recommendation

div2vec: Diversity-Emphasized Node Embedding