KDD2020推荐系统论文聚焦

前言

ACM SIGKDD（国际数据挖掘与知识发现大会，简称KDD）是数据挖掘领域的最高级别的学术会议，由ACM的数据挖掘及知识发现专委会（SIGKDD）主办，是CCF A类会议。其中今年的Research Track接受率为216/1279=16.8%。

推荐系统论文列表

这次整理的推荐系统论文列表分为了Research Track和Applied Data Science Track，即面向研究型的学术论文和面向工业界的实践论文。

Research Track Papers

研究赛道的论文主要是按照推荐子领域来划分，比如序列化推荐、对话推荐系统、冷启动问题、协同过滤、推荐效率问题等。从以下比例可以看出，序列化推荐和对话推荐系统是研究的热点问题，这其实也很容易理解，推荐其实是个天然的序列问题，即建模用户的一系列行为同时返回一系列个性化的物品序列；同时，推荐系统也自然的引入对话机制，因为传统的推荐是静态的，用户只能被动的接受着推荐系统返回的结果列表，引入对话交互机制后，能很好的优化推荐系统。

Sequential RS

• Disentangled Self-Supervision in Sequential Recommenders

• Counterfactual Evaluation of Slate Recommendations with Sequential Reward Interactions

• Geography-Aware Sequential Location Recommendation

• Handling Information Loss of Graph Neural Networks for Session-based Recommendation

• On Sampling Top-K Recommendation Evaluation

• Probabilistic Metric Learning with Adaptive Margin for Top-K Recommendation

Conversational RS

• Evaluating Conversational Recommender Systems via User Simulation

• Improving Conversational Recommender Systems via Knowledge Graph based Semantic Fusion

• Interactive Path Reasoning on Graph for Conversational Recommendation

Cold-Start in RS

• MAMO: Memory-Augmented Meta-Optimization for Cold-start Recommendation

• Meta-learning on Heterogeneous Information Networks for Cold-start Recommendation

Collaborative Filtering

• Dual Channel Hypergraph Collaborative Filtering

• Semi-supervised Collaborative Filtering by Text-enhanced Domain Adaptation

Efficient RS

• An Efficient Neighborhood-based Interaction Model for Recommendation on Heterogeneous Graph

• Compositional Embeddings Using Complementary Partitions for Memory-Efficient Recommendation Systems

• FedFast: Going Beyond Average for Faster Training of Federated Recommender Systems

Others

• A Framework for Recommending Accurate and Diverse Items Using Bayesian Graph Convolutional Neural Networks

• BLOB: A Probabilistic Model for Recommendation that Combines Organic and Bandit Signals

• Joint Policy-Value Learning for Recommendation

• On Sampled Metrics for Item Recommendation

Applied Data Science Track Papers

应用数据科学赛道主要是展示工业界中的实践成果，我们按照公司维度整理出了涉及推荐场景的论文，其中包括谷歌、阿里、亚马逊等公司，这些公司由于有着海量的用户数据，因此推荐技术也相对成熟，许多经典模型也是由以下公司所提出的。

Google

• Attribute-based Propensity for Unbiased Learning in Recommender Systems: Algorithm and Case Studies

• Improving Recommendation Quality in Google Drive

• Neural Input Search for Large Scale Recommendation Models

Alibaba

• Controllable Multi-Interest Framework for Recommendation

• M2GRL: A Multi-task Multi-view Graph Representation Learning Framework for Web-scale Recommender Systems

• Maximizing Cumulative User Engagement in Sequential Recommendation: An Online Optimization Perspective

• Privileged Features Distillation at Taobao Recommendations -Alibaba

Amazon

• Temporal-Contextual Recommendation in Real-Time

Pinterest

• PinnerSage: Multi-Modal User Embedding Framework for Recommendations at Pinterest -Pinterest

Bytedance

• Jointly Learning to Recommend and Advertise -Bytedance

DiDiChuxing

• Gemini: A novel and universal heterogeneous graph information fusing framework for online recommendations

Twitter

• SimClusters: Community-Based Representations for Heterogeneous Recommendations at Twitter

更多论文，欢迎访问https://github.com/hongleizhang/RSPapers