1. 推荐系统强化学习研究综述

Title: A Survey on Reinforcement Learning for Recommender Systems

Published: 2022-03-24

Url: http://arxiv.org/abs/2109.10665v2

Authors: Yuanguo Lin,Yong Liu,Fan Lin,Lixin Zou,Pengcheng Wu,Wenhua Zeng,Huanhuan Chen,Chunyan Miao

推荐系统被广泛应用于不同的现实生活场景，帮助我们找到有用的信息。特别是，基于强化学习（RL）的推荐系统近年来已成为一个新兴的研究课题。实验结果表明，基于RL的推荐方法由于其交互性和自主学习能力，在监督学习方法中占绝大多数。然而，在推荐系统中应用RL有各种各样的挑战。为了理解这些挑战和相关的解决方案，研究人员和实践者可以参考基于RL的推荐系统。为此，我们首先对RL方法在典型推荐场景中的应用进行了全面的概述、比较和总结，包括交互式推荐、会话推荐、顺序推荐和解释性推荐。在现有文献的基础上，我们进一步系统地分析了挑战和相关解决方案。最后，在讨论RL的开放性问题及其推荐系统的局限性的基础上，我们强调了该领域的一些潜在研究方向。

Recommender systems have been widely applied in different real-life scenariosto help us find useful information. In particular, Reinforcement Learning (RL)based recommender systems have become an emerging research topic in recentyears. Empirical results show that RL-based recommendation methods oftensurpass most of supervised learning methods, owing to the interactive natureand autonomous learning ability. Nevertheless, there are various challenges ofapplying RL in recommender systems. To understand the challenges and relevantsolutions, there should be a reference for researchers and practitionersworking on RL-based recommender systems. To this end, we firstly provide athorough overview, comparisons, and summarization of RL approaches applied infour typical recommendation scenarios, including interactive recommendation,conversational recommendatin, sequential recommendation, and explainablerecommendation. We further systematically analyze the challenges and relevantsolutions on the basis of existing literature. Finally, under discussion foropen issues of RL and its limitations of recommender systems, we highlight somepotential research directions in this field.

2. 衡量推荐系统中的“为什么”：可解释推荐评估综述

Title: Measuring "Why" in Recommender Systems: a Comprehensive Survey on the Evaluation of Explainable Recommendation

Published: 2022-02-14

Url: http://arxiv.org/abs/2202.06466v1

Authors: Xu Chen,Yongfeng Zhang,Ji-Rong Wen

解释性推荐在提高推荐说服力、用户满意度、系统透明度等方面显示出了巨大的优势。解释性推荐的一个基本问题是如何评估解释。在过去几年中，人们提出了各种评估策略。然而，它们分散在不同的论文中，缺乏系统和详细的比较。为了弥合这一差距，本文全面回顾了前人的工作，并根据评估视角和评估方法为他们提供了不同的建议。除了总结之前的工作之外，我们还分析了现有评估方法的优点，并提供了一系列关于如何选择它们的指南。这项调查的内容基于来自IJCAI、AAAI、TheWebConf、Recsys、UMAP和IUI等顶级会议的100多篇论文，并在石墨文档中给出了它们的完整摘要-https://shimo.im/sheets/VKrpYTcwVH6KXgdy/MODOC/.通过这次调查，我们最终希望对可解释性建议的评估提供一个清晰而全面的回顾。

Explainable recommendation has shown its great advantages for improvingrecommendation persuasiveness, user satisfaction, system transparency, amongothers. A fundamental problem of explainable recommendation is how to evaluatethe explanations. In the past few years, various evaluation strategies havebeen proposed. However, they are scattered in different papers, and there lacksa systematic and detailed comparison between them. To bridge this gap, in thispaper, we comprehensively review the previous work, and provide differenttaxonomies for them according to the evaluation perspectives and evaluationmethods. Beyond summarizing the previous work, we also analyze the(dis)advantages of existing evaluation methods and provide a series ofguidelines on how to select them. The contents of this survey are based on morethan 100 papers from top-tier conferences like IJCAI, AAAI, TheWebConf, Recsys,UMAP, and IUI, and their complete summarization are presented at https://shimo.im/sheets/VKrpYTcwVH6KXgdy/MODOC/. With this survey, we finallyaim to provide a clear and comprehensive review on the evaluation ofexplainable recommendation.

3. 推荐系统中的图神经网络：综述

Title: Graph Neural Networks in Recommender Systems: A Survey

Published: 2022-02-13

Url: http://arxiv.org/abs/2011.02260v3

Authors: Shiwen Wu,Fei Sun,Wentao Zhang,Xu Xie,Bin Cui

随着在线信息的爆炸式增长，推荐系统在缓解此类信息过载方面发挥着关键作用。由于推荐系统的重要应用价值，这一领域的研究一直是新兴的。在推荐系统中，主要的挑战是从他们的交互和附带信息（如果有的话）中学习有效的用户/项目表示。近年来，由于推荐系统中的大部分信息本质上都是图结构，且GNN在表示学习方面具有优越性，因此图形神经网络（graph neural network，GNN）技术在推荐系统中得到了广泛的应用。本文旨在全面介绍基于GNN的推荐系统的最新研究成果。具体来说，我们根据使用的信息类型和推荐任务，提供了基于GNN的推荐模型的分类。此外，我们还系统地分析了在不同类型的数据上应用GNN所面临的挑战，并讨论了该领域现有的工作如何应对这些挑战。此外，我们陈述了与该领域发展相关的新观点。我们收集了具有代表性的论文及其开源实现https://github.com/wusw14/GNN-in-RS.

With the explosive growth of online information, recommender systems play akey role to alleviate such information overload. Due to the importantapplication value of recommender systems, there have always been emerging worksin this field. In recommender systems, the main challenge is to learn theeffective user/item representations from their interactions and sideinformation (if any). Recently, graph neural network (GNN) techniques have beenwidely utilized in recommender systems since most of the information inrecommender systems essentially has graph structure and GNN has superiority ingraph representation learning. This article aims to provide a comprehensivereview of recent research efforts on GNN-based recommender systems.Specifically, we provide a taxonomy of GNN-based recommendation modelsaccording to the types of information used and recommendation tasks. Moreover,we systematically analyze the challenges of applying GNN on different types ofdata and discuss how existing works in this field address these challenges.Furthermore, we state new perspectives pertaining to the development of thisfield. We collect the representative papers along with their open-sourceimplementations in https://github.com/wusw14/GNN-in-RS.