推荐系统作为深度学习御三家（CV, NLP, RS）之一，一直都是学术界和工业界的热门研究 topic。为了更加清楚的掌握推荐系统的前沿方向与最新进展，本文整理了最近一年顶会中推荐系统相关的论文，一共涵盖 SIGIR2020, KDD2020, RecSys2020, CIKM2020, AAAI2021, WSDM2021, WWW2021 七个会议共 221 篇论文。本次整理以 long paper 和 research paper 为主，也包含少量的 short paper 和 industry paper。
1. Sequential Recommendation with Self-attentive Multi-adversarial Network. SIGIR 20202. KERL: A Knowledge-Guided Reinforcement Learning Model for Sequential Recommendation. SIGIR 2020 【SR + KG + RL】3. Modeling Personalized Item Frequency Information for Next-basket Recommendation. SIGIR 2020 【融合item频率】4. Incorporating User Micro-behaviors and Item Knowledge into Multi-task Learning for Session-based Recommendation. SIGIR 2020【SR + KG + MTL】5. GAG: Global Attributed Graph Neural Network for Streaming Session-basedRecommendation. SIGIR 2020 【Streaming SR】6. Next-item Recommendation with Sequential Hypergraphs. SIGIR 2020 【基于超图】7. A General Network Compression Framework for Sequential Recommender Systems. SIGIR 2020 【通用的SR模型压缩方法】8. Make It a Chorus: Knowledge- and Time-aware Item Modeling for Sequential Recommendation. SIGIR 2020 【SR + KG】9. Global Context Enhanced Graph Neural Networks for Session-based Recommendation. SIGIR 202010. Self-Supervised Reinforcement Learning for Recommender Systems. SIGIR 2020 【SR + RL】11. Time Matters: Sequential Recommendation with Complex Temporal Information. SIGIR 2020 【Time-aware】12. Controllable Multi-Interest Framework for Recommendation. SIGIR 2020 【多兴趣】13. Disentangled Self-Supervision in Sequential Recommenders. KDD 2020 【多兴趣】14. Handling Information Loss of Graph Neural Networks for Session-based Recommendation. KDD 202015. Contextual and Sequential User Embeddings for Large-Scale Music Recommendation. RecSys 2020 【针对music场景】16. FISSA:Fusing Item Similarity Models with Self-Attention Networks for Sequential Recommendation. RecSys 2020 【SASRec的基础上融合物品相似性】17. From the lab to production: A case study of session-based recommendations in the home-improvement domain. RecSys 2020 【对SR方法评测分析】18. Recommending the Video to Watch Next: An Offline and Online Evaluation at YOUTV.de. RecSys 2020 【基于Markov Chain】19. SSE-PT:Sequential Recommendation Via Personalized Transformer. RecSys 2020 【改进SASRec】20. Improving End-to-End Sequential Recommendations with Intent-aware Diversification. CIKM 202021. Quaternion-based self-Attentive Long Short-term User Preference Encoding for Recommendation. CIKM 2020 【Quaternion捕捉长短期兴趣】22. Sequential Recommender via Time-aware Attentive Memory Network. CIKM 2020 【Time-aware】23. Star Graph Neural Networks for Session-based Recommendation. CIKM 202024. Dynamic Memory Based Attention Network for Sequential Recommendation. AAAI 202125. Noninvasive Self-Attention for Side Information Fusion in Sequential Recommendation. AAAI 2021【BERT + Context】26. Self-Supervised Hypergraph Convolutional Networks for Session-based Recommendation. AAAI 2021 【基于超图】27. An Efficient and Effective Framework for Session-based Social Recommendation. WSDM 202128. Sparse-Interest Network for Sequential Recommendation. WSDM 2021 【多兴趣】29. Dynamic Embeddings for Interaction Prediction. WWW 202130. Session-aware Linear Item-Item Models for Session-based Recommendation. WWW 202131. RetaGNN: Relational Temporal Attentive Graph Neural Networks for Holistic Sequential Recommendation. WWW 202132. Adversarial and Contrastive Variational Autoencoder for Sequential Recommendation. WWW 2021 【VAE】33. Future-Aware Diverse Trends Framework for Recommendation. WWW 2021 【融合特征】34. Linear-Time Self Attention with Codeword Histogram for Efficient Recommendation. WWW 2021 【线性时间的selfattention】35. DeepRec: On-device Deep Learning for Privacy-Preserving Sequential Recommendation in MobileCommerce. WWW 2021 【序列推荐中的隐私保护】
1.3 Knowledge-aware Recommendations
这一部分主要包括利用结构化信息来帮助推荐系统的工作，用到的结构化信息主要以 Knowledgegraph 和 HIN 为主。这一类推荐任务也是近些年来热门的研究方向。为了能充分发挥出结构化信息的优势，早期有不少基于 meta-path 的方法，随着图神经网络的发展，目前 GNN 的方法已经成为主流。1. CKAN: Collaborative Knowledge-aware Attentive Network for Recommender Systems. SIGIR 20202. Attentional Graph Convolutional Networks for Knowledge Concept Recommendation in MOOCs in a Heterogeneous View. SIGIR2020 【MOOC推荐】3. MVIN: Learning multiview items for recommendation. SIGIR 20204. Jointly Non-Sampling Learning for Knowledge Graph Enhanced Recommendation. SIGIR 20205. Joint Item Recommendation and Attribute Inference: An Adaptive Graph Convolutional Network Approach. SIGIR 20206. Leveraging Demonstrations for Reinforcement Recommendation Reasoning over Knowledge Graphs. SIGIR 2020 【RL】7. SimClusters Community-Based Representations for Heterogenous Recommendations at Twitter. KDD 2020 【IndustryPaper by Twitter】8. Multi-modal Knowledge Graphs for Recommender Systems. CIKM 20209. DisenHAN Disentangled Heterogeneous Graph Attention Network for Recommendation. CIKM 202010. Genetic Meta-Structure Search for Recommendation on Heterogeneous Information Network. CIKM 2020 【自动优化meta-path】11. TGCN Tag Graph Convolutional Network for Tag-Aware Recommendation. CIKM 202012. Knowledge-Enhanced Top-K Recommendation in Poincaré Ball. AAAI 202113. Graph Heterogeneous Multi-Relational Recommendation. AAAI 202114. Knowledge-Enhanced Hierarchical Graph Transformer Network for Multi-Behavior Recommendation. AAAI 202115. Alleviating Cold-Start Problems in Recommendation through Pseudo-Labelling over Knowledge Graph. WSDM 202116. Decomposed Collaborative Filtering Modeling Explicit and Implicit Factors For Recommender Systems. WSDM 202117. Temporal Meta-path Guided Explainable Recommendation. WSDM 202118. Learning Intents behind Interactions with Knowledge Graph for Recommendation. WWW 2021
1.4 Feature Interactions
这一部分工作主要以特征交互为主。已经有非常多经典的方法在工业界的推荐中得到了广泛的应用，例如 FM，DeepFM，Wide&Deep 等。1. Detecting Beneficial Feature Interactions for Recommender Systems. AAAI 2021 【Graph-based】2. DeepLight: Deep Lightweight Feature Interactions for Accelerating CTR Predictions in Ad Serving. WSDM 2021 【加速特征交互】3. Multi-Interactive Attention Network for Fine-grained Feature Learning in CTR Prediction. WSDM 2021 【细粒度】4. FM^2: Field-matrixed Factorization Machines for CTR Prediction. WWW 2021 【FwFM升级版】
1.5 Conversational Recommender System
1. Towards Question-based Recommender Systems.SIGIR 20202. Improving Conversational Recommender Systems via Knowledge Graph based Semantic Fusion. KDD 2020 【CRS + KG】3. Interactive Path Reasoning on Graph for Conversational Recommendation. KDD 20204. A Ranking Optimization Approach to Latent Linear Critiquing for Conversational Recommender Systems. RecSys 2020 【对话推荐系统的一种排序优化方法】5. What does BERT know about books, movies and music：Probing BERT for Conversational Recommendation.RecSys 2020 【CRS + KG】6. Adapting User Preference to Online Feedback in Multi-round Conversational Recommendation. WSDM 2021 【多轮】7. A Workflow Analysis of Context-driven Conversational Recommendation. WWW 2021 【分析CRS workflow】
1.6 Social Recommendations
1. Partial Relationship Aware Influence Diffusion via a Multi-channel Encoding Scheme for Social Recommendation. CIKM20202. Random Walks with Erasure: Diversifying Personalized Recommendations on Social and Information Networks. WWW 20213. Dual Side Deep Context-aware Modulation for Social Recommendation. WWW 20214. Self-Supervised Multi-Channel Hypergraph Convolutional Network for Social Recommendation. WWW 2021
1.7 News Recommendations
1. KRED: Knowledge-Aware Document Representation for News Recommendations. RecSys 20202. News Recommendation with Topic-Enriched Knowledge Graphs. CIKM 20203. The Interaction between Political Typology and Filter Bubbles in News Recommendation Algorithms. WWW 2021
1.8 Text-aware Recommendations
1. TAFA: Two-headed Attention Fused Autoencoder for Context-Aware Recommendations. RecSys 2020 【Review】2. Set-Sequence-Graph A Multi-View Approach Towards Exploiting Reviews for Recommendation. CIKM 2020 【Review】3. TPR: Text-aware Preference Ranking for Recommender Systems. CIKM 20204. Leveraging Review Properties for Effective Recommendation. WWW 2021 【Review】
1.9 Point-of-Interest
1. HME: A Hyperbolic Metric Embedding Approach for Next-POI Recommendation. SIGIR 20202. Spatial Object Recommendation with Hints: When Spatial Granularity Matters. SIGIR 20203. Geography-Aware Sequential Location Recommendation. KDD 20204. Learning Graph-Based Geographical Latent Representation for Point-of-Interest Recommendation. CIKM 20205. STP-UDGAT Spatial-Temporal-Preference User Dimensional Graph Attention Network for Next POI Recommendation. CIKM 20206. STAN: Spatio-Temporal Attention Network for next Point-of-Interest Recommendation. WWW 20217. Incremental Spatio-Temporal Graph Learning for Online Query-POI Matching. WWW 2021
1.10 Online Recommendations
1. Gemini: A novel and universal heterogeneous graph information fusing framework for online recommendations. KDD 20202. Maximizing Cumulative User Engagement in Sequential Recommendation An Online Optimization Perspective. KDD 20203. Exploring Clustering of Bandits for Online Recommendation System. RecSys 20204. Contextual User Browsing Bandits for Large-Scale Online Mobile Recommendation. RecSys 20205. A Hybrid Bandit Framework for Diversified Recommendation. AAAI 2021
1.11Group Recommendations
1. GAME: Learning Graphical and Attentive Multi-view Embeddings for Occasional Group Recommendation. SIGIR 20202. GroupIM: A Mutual Information Maximizing Framework for Neural Group Recommendation. SIGIR 20203. Group-Aware Long- and Short-Term Graph Representation Learning for Sequential Group Recommendation. SIGIR 2020
1.12 Multi-task/Multi-behavior/Cross-domain
1. Transfer Learning via Contextual Invariants for One-to-Many Cross-Domain Recommendation. SIGIR 2020 【Cross-domain】2. CATN: Cross-Domain Recommendation for Cold-Start Users via Aspect Transfer Network. SIGIR 2020 【Cross-domain】3. Multi-behavior Recommendation with Graph Convolution Networks. SIGIR 20204. Parameter-Efficient Transfer from Sequential Behaviors for User Modeling and Recommendation. SIGIR 20205. Web-to-Voice Transfer for Product Recommendation on Voice. SIGIR 20206. Jointly Learning to Recommend andAdvertise. KDD 20207. Progressive Layered Extraction (PLE) A Novel Multi-Task Learning (MTL) Model for Personalized Recommendations. RecSys2020 【多目标优化 by Tencent】8. Whole-Chain Recommendations. CIKM 20209. Personalized Approximate Pareto-Efficient Recommendation. WWW 2021 【面向Pareto的强化学习方法解决多目标优化推荐】
1.13 Other Task
除了上面提到的比较经典的推荐任务，最近一年的顶会文章还有一些其他有意思的任务。1. Hierarchical Fashion Graph Network for Personalized Outfit Recommendation. SIGIR 2020 【OutfitRecommendation】2. Octopus: Comprehensive and Elastic User Representation for the Generation of Recommendation Candidates. SIGIR 2020 【Candidate Generation】3. Goal-driven Command Recommendations for Analysts. RecSys 2020 【从非结构化log数据中进行command recommendation】4. MultiRec: A Multi-Relational Approach for Unique Item Recommendation in Auction Systems. RecSys 2020 【拍卖系统中的推荐】5. PURS: Personalized Unexpected Recommender System for Improving User Satisfaction. RecSys 2020 【意外推荐，没想到吧.jpg】6. RecSeats: A Hybrid Convolutional Neural Network Choice Model for Seat 7. Recommendations at Reserved Seating Venues.RecSys 2020 【座位推荐，莱纳你坐啊】8. Live Multi-Streaming and Donation Recommendations via Coupled Donation-Response Tensor Factorization. CIKM 2020 【Multi-streaming】9. Learning to Recommend from Sparse Data via Generative User Feedback. AAAI 202110. Real-time Relevant Recommendation Suggestion. WSDM 2021 【Recommendation Suggestion】11. Heterogeneous Graph Augmented Multi-Scenario Sharing Recommendation with Tree-Guided Expert Networks. WSDM2021 【Share Recommendation】12. FINN: Feedback Interactive Neural Network for Intent Recommendation. WWW 2021 【IntentRecommendation】13. Drug Package Recommendation via Interaction-aware Graph Induction. WWW 2021 【DrugPackage Recommendation】14. Large-scale Comb-K Recommendation. WWW 2021【Comb-K 推荐】15. Variation Control and Evaluation for Generative Slate Recommendations. WWW 2021 【GenerativeSlate Recommendation】16. Diversified Recommendation Through Similarity-Guided Graph Neural Networks. WWW 2021 【多样性推荐】
02
推荐的热门研究话题
2.1 Debias in Recommender System
bias 是广泛存在于推荐系统中的，包括比较常见的流行度偏差 (popularity bias)，选择偏差 (selection bias)，曝光偏差 (exposure bias)，位置偏差 (position bias) 以及其他各种偏差。推荐系统的 debias 的工作一直都有人在做，随着近两年 causal inference 成为热点，这个 topic 最近变得非常热门。1. A General Knowledge Distillation Framework for Counterfactual Recommendation via Uniform Data. SIGIR 20202. Measuring and Mitigating Item Under-Recommendation Bias in Personalized Ranking Systems. SIGIR 20203. Attribute-based Propensity for Unbiased Learning in Recommender Systems Algorithm and Case Studies. KDD 2020 【position bias】4. Counterfactual Evaluation of Slate Recommendations with Sequential Reward Interactions. KDD 20205. Debiasing Item-to-Item Recommendations With Small Annotated Datasets. RecSys 20206. Keeping Dataset Biases out of the Simulation : A Debiased Simulator for Reinforcement Learning based RecommenderSystems. RecSys 20207. Unbiased Ad Click Prediction for Position-aware Advertising Systems. RecSys 2020 【debiasin position-aware recommedantion】8. Unbiased Learning for the Causal Effect of Recommendation. RecSys 20209. E-commerce Recommendation with Weighted Expected Utility. CIKM 202010. Popularity-Opportunity Bias in Collaborative Filtering. WSDM 2021 【提出了一种新的popularitybias考虑了opportunity】11. Combating Selection Biases in Recommender Systems with a Few Unbiased Ratings. WSDM 2021 【利用少部分无偏数据解决selection bias】12. Leave No User Behind Towards Improving the Utility of Recommender Systems for Non-mainstream Users. WSDM 2021 【mainstream bias】13. Non-Clicks Mean Irrelevant Propensity Ratio Scoring As a Correction. WSDM 202114. Diverse User Preference Elicitation with Multi-Armed Bandits. WSDM 2021 【多臂老虎机增加推荐系统多样性缓解popularitybias】15. Unbiased Learning to Rank in Feeds Recommendation. WSDM 2021 【context-aware position bias】16. Cross-Positional Attention for Debiasing Clicks. WWW 202117. Debiasing Career Recommendations with Neural Fair Collaborative Filtering. WWW 2021 【genderbias】
2.2 Fairness in Recommender System
1. airness-Aware Explainable Recommendation over Knowledge Graphs. SIGIR 2020 【运用KG做re-ranking】2. Ensuring Fairness in Group Recommendations by Rank-Sensitive Balancing of Relevance. RecSys 20203. Fairness-Aware News Recommendation with Decomposed Adversarial Learning. AAAI 2021 【Fairness inNews Recommendation】4. Practical Compositional Fairness Understanding Fairness in Multi-Component Recommender Systems. WSDM 20215. Towards Long-term Fairness in Recommendation. WSDM 20216. Learning Fair Representations for Recommendation: A Graph-based Perspective. WWW 20217. User-oriented Group Fairness In Recommender Systems. WWW 2021
2.3 Attack in Recommender System
万物皆可攻击：1. Revisiting Adversarially Learned Injection Attacks Against Recommender Systems. RecSys 2020 【对推荐系统对抗性注入攻击的重新审视】2. Attacking Recommender Systems with Augmented User Profiles. CIKM 2020 【从用户特征进行攻击】3. A Black-Box Attack Model for Visually-Aware Recommenders. WSDM 2021 【针对图像特征的推荐系统攻击】4. Denoising Implicit Feedback for Recommendation. WSDM 2021 【训练数据去噪】5. Adversarial Item Promotion: Vulnerabilities at the Core of Top-N 6. Recommenders that Use Images to Address Cold Start. WWW2021 【针对图像攻击】6. Graph Embedding for Recommendation against Attribute Inference Attacks. WWW 2021
2.4 Explanation in Recommender System
1. Try This Instead: Personalized and Interpretable Substitute Recommendation. KDD 20202. CAFE: Coarse-to-Fine Neural Symbolic Reasoning for Explainable Recommendation. CIKM 2020 【通过KG提升可解释性】3. Explainable Recommender Systems via Resolving Learning Representations. CIKM 2020 【借助KG表示学习提升可解释性】4. Generate Neural Template Explanations for Recommendation. CIKM 2020 【生成文本提升可解释性】5. Explainable Recommendation with Comparative Constraints on Product Aspects. WSDM 2021 【通过和物品在属性上比较来提升可解释性】6. Explanation as a Defense of Recommendation. WSDM 2021 【生成文本提升可解释性】7. EX^3: Explainable Product Set Recommendation for Comparison Shopping. WWW 20218. Learning from User Feedback on Explanations to Improve Recommender Models. WWW 2021
2.5 Long-tail/Cold-start in Recommender System
从有推荐系统开始就存在的痛点问题，目前比较流行基于 meta-learning, transfer-learning 的方法。1. Content-aware Neural Hashing for Cold-start Recommendation. SIGIR 20202. MAMO: Memory-Augmented Meta-Optimization for Cold-start Recommendation. KDD 20203. Learning Transferrable Parameters for Long-tailed Sequential User Behavior Modeling. KDD 2020 【通用的方法解决序列推荐中的长尾问题】4. Meta-learning on Heterogeneous Information Networks for Cold-start Recommendation. KDD 20205. Cold-Start Sequential Recommendation via Meta Learner. AAAI 20216. Personalized Adaptive Meta Learning for Cold-start User Preference Prediction. AAAI 20217. Task-adaptive Neural Process for User Cold-Start Recommendation. WWW 20218. A Model of Two Tales: Dual Transfer Learning Framework for Improved Long-tail Item Recommendation. WWW 2021
2.6 Evaluation
对于新型推荐任务评测方式的创造以及到底该不该采样评测？1. Measuring Recommendation Explanation Quality: The Conflicting Goals of Explanations. SIGIR 2020 【评价生成的推荐系统解释的质量】2. Evaluating Conversational Recommender Systems via User Simulation. KDD 2020 【CRS评测指标】3. On Sampled Metrics for Item Recommendation.KDD 20204. On Sampling Top-K Recommendation Evaluation. KDD 20205. Are We Evaluating Rigorously:Benchmarking Recommendation for Reproducible Evaluation and FairComparison. RecSys 20206. On Target Item Sampling in Offline Recommender System Evaluation. RecSys 2020
03
先进技术在推荐中的应用
3.1 Pre-training in Recommender System
1. S^3-Rec: Self-Supervised Learning for Sequential Recommendation with Mutual Information Maximization. CIKM 20202. U-BERT Pre-Training User Representations for Improved Recommendation. AAAI 20213. Pre-Training Graph Neural Networks for Cold-Start Users and Items Representation. WSDM 2021
3.2 Reinforcement Learning in Recommender System
1. MaHRL: Multi-goals Abstraction based Deep Hierarchical Reinforcement Learning for Recommendations. SIGIR 20202. Interactive Recommender System via Knowledge Graph-enhanced Reinforcement Learning. SIGIR 20203. Joint Policy-Value Learning for Recommendation. KDD 20204. BLOB: A Probabilistic Model for Recommendation that Combines Organic and Bandit Signals. KDD 20205. Learning to Collaborate in Multi-Module Recommendation via Multi-Agent Reinforcement Learning without Communication.RecSys 20206. Reinforcement Learning with a Disentangled Universal Value Function for Item Recommendation. AAAI 20217. User Response Models to Improve a REINFORCE Recommender System. WSDM 20218. Cost-Effective and Interpretable Job Skill Recommendation with Deep Reinforcement Learning. WWW 20219. A Multi-Agent Reinforcement Learning Framework for Intelligent Electric Vehicle Charging Recommendation. WWW 202110. Reinforcement Recommendation with User Multi-aspect Preference. WWW 2021
3.3 Knowledge Distillation in Recommender System
1. Privileged Features Distillation at Taobao Recommendations. KDD 20202. DE-RRD: A Knowledge Distillation Framework for Recommender System. CIKM 2020 【BPRMF也需要蒸馏我是没想到的】3. Bidirectional Distillation for Top-K Recommender System. WWW 2021 【双向蒸馏】
3.4 NAS in Recommender System
1. Neural Input Search for Large Scale Recommendation Models. KDD 20202. Field-aware Embedding Space Searching in Recommender Systems. WWW 2021 【AutoML自动选择特征维度】
3.5 Federated Learning in Recommender System
1. FedFast Going Beyond Average for Faster Training of Federated Recommender Systems. KDD 2020
04
理论/实验分析
推荐系统理论分析的论文不是很常见，下面这几篇都还挺有趣的。1. How Dataset Characteristics Affect the Robustness of Collaborative Recommendation Models. SIGIR 20202. Agreement and Disagreement between True and False-Positive Metrics in Recommender Systems Evaluation. SIGIR 20203. Critically Examining the Claimed Value of Convolutions over User-Item Embedding Maps for Recommender Systems. CIKM 2020 【理论结合实验分析CNN建模推荐系统embedding不太work】4. On Estimating Recommendation Evaluation Metrics under Sampling. AAAI 20215. Beyond Point Estimate Inferring Ensemble Prediction Variation from Neuron Activation Strength in Recommender Systems.WSDM 2021 【分析推荐系统ensemble】6. Bias-Variance Decomposition for Ranking. WSDM 20217. Theoretical Understandings of Product Embedding for E-commerce Machine Learning. WSDM 2021
05
其他
1. Learning Personalized Risk Preferences for Recommendation. SIGIR 20202. Distributed Equivalent Substitution Training for Large-Scale Recommender Systems. SIGIR 2020 【大规模推荐系统的分布式训练】3. Beyond User Embedding Matrix: Learning to Hash for Modeling Large-Scale Users in Recommendation. SIGIR 2020 【大规模user hash】4. How to Retrain a Recommender System? SIGIR2020 【来了新数据怎么retrain】5. Recommendation for New Users and New Items via Randomized Training and Mixture-of-Experts Transformation. SIGIR 20206. Compositional Embeddings Using Complementary Partitions for Memory-Efficient Recommendation Systems. KDD 2020 【解决embedding内存问题】7. Improving Recommendation Quality in Google Drive. KDD 2020 【Google Drive 推荐实战】8. A Method to Anonymize Business Metrics to Publishing Implicit Feedback Datasets. RecSys 2020 【如何构建和发布数据集】9. Exploiting Performance Estimates for Augmenting Recommendation Ensembles. RecSys 2020 【有效的推荐模型ensemble方法】10. User Simulation via Supervised Generative Adversarial Network. WWW 2021
🔍
现在，在「知乎」也能找到我们了
进入知乎首页搜索「PaperWeekly」
点击「关注」订阅我们的专栏吧
关于PaperWeekly
PaperWeekly 是一个推荐、解读、讨论、报道人工智能前沿论文成果的学术平台。如果你研究或从事 AI 领域，欢迎在公众号后台点击「交流群」，小助手将把你带入 PaperWeekly 的交流群里。