推薦算法最前沿|KDD2020推薦系統論文一覽

作者：學派

連結：https://zhuanlan.zhihu.com/p/161705748

KDD（https://www.kdd.org/kdd2020/）是推薦領域一個頂級的國際會議。本次接收的論文按照推薦系統應用場景可以大緻劃分為：CTR預估、TopN推薦、對話式推薦、序列推薦等。同時，GNN、強化學習、多任務學習、遷移學習、AutoML、元學習在推薦系統的落地應用也成為當下的主要研究點。此屆會議有很大一部分來自工業界的論文，包括Google、Microsoft、Criteo、Spotify以及國内大廠阿裡、百度、位元組、華為、滴滴等。

CTR Prediction

1. AutoFIS: Automatic Feature Interaction Selection in Factorization Models for Click-Through Rate Prediction 【華為諾亞】

簡介：本文采用AutoML的搜尋方法選擇重要性高的二次特征互動項、去除幹擾項，提升FM、DeepFM這類模型的準确率。

論文：arxiv.org/abs/2003.1123

2. Category-Specific CNN for Visual-aware CTR Prediction at JD.com 【京東】

論文：arxiv.org/abs/2006.1033

3. Towards Automated Neural Interaction Discovering for Click-Through Rate Prediction 【Facebook】

論文：arxiv.org/abs/2007.0643

Graph-based Recommendation

1. A Framework for Recommending Accurate and Diverse Items Using Bayesian Graph Convolutional Neural Networks 【華為諾亞】

2. An Efficient Neighborhood-based Interaction Model for Recommendation on Heterogeneous Graph 【Amazon】

論文：arxiv.org/abs/2007.0021

3. M2GRL: A Multi-task Multi-view Graph Representation Learning Framework for Web-scale Recommender Systems 【阿裡】

簡介：本文通過關聯多個視角的圖(item-item圖、item-shop圖、shop-shop圖等)增強item表征，用于item召回。

論文：arxiv.org/abs/2005.1011

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

5. Interactive Path Reasoning on Graph for Conversational Recommendation

論文：arxiv.org/abs/2007.0019

6. A Dual Heterogeneous Graph Attention Network to Improve Long-Tail Performance for Shop Search in E-Commerce 【阿裡】

7. Gemini: A Novel and Universal Heterogeneous Graph Information Fusing Framework for Online Recommendations 【滴滴】

Conversational Recommendation

1. Evaluating Conversational Recommender Systems via User Simulation

論文：arxiv.org/abs/2006.0873

2. Improving Conversational Recommender Systems via Knowledge Graph based Semantic Fusion

論文：arxiv.org/abs/2007.0403

3. Interactive Path Reasoning on Graph for Conversational Recommendation

論文：arxiv.org/abs/2007.0019

CF and Top-N Recommendation

1. Dual Channel Hypergraph Collaborative Filtering 【百度】

筆記：blog.csdn.net/weixin_42

2. Probabilistic Metric Learning with Adaptive Margin for Top-K Recommendation 【華為諾亞】

3. Controllable Multi-Interest Framework for Recommendation 【阿裡】

論文：arxiv.org/abs/2005.0934

4. Embedding-based Retrieval in Facebook Search 【Facebook】

論文：arxiv.org/abs/2006.1163

5. On Sampling Top-K Recommendation Evaluation

Embedding and Representation

1. Compositional Embeddings Using Complementary Partitions for Memory-Efficient Recommendation Systems 【Facebook】

論文：arxiv.org/abs/1909.0210

2. PinnerSage: Multi-Modal User Embedding Framework for Recommendations at Pinterest 【Pinterest】

論文：arxiv.org/abs/2007.0363

3. SimClusters: Community-Based Representations for Heterogeneous Recommendations at Twitter 【Twitter】

4. Time-Aware User Embeddings as a Service 【Yahoo】

論文：astro.temple.edu/~tuf28

Sequential Recommendation

1. Disentangled Self-Supervision in Sequential Recommenders 【阿裡】

論文：http://pengcui.thumedialab.com/papers/Disen...

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

3. Maximizing Cumulative User Engagement in Sequential Recommendation: An Online Optimization Perspective 【阿裡】

論文：arxiv.org/pdf/2006.0452

RL for Recommendation

1. Jointly Learning to Recommend and Advertise 【位元組跳動】

論文：arxiv.org/abs/2003.0009

2. BLOB: A Probabilistic Model for Recommendation that Combines Organic and Bandit Signals 【Criteo】

3. Joint Policy-Value Learning for Recommendation 【Criteo】

論文：researchgate.net/public

Multi-Task Learning

1. Privileged Features Distillation at Taobao Recommendations 【阿裡】

論文：arxiv.org/abs/1907.0517

Transfer Learning

1. Learning Transferrable Parameters for Long-tailed Sequential User Behavior Modeling 【Salesforce】

2. Semi-supervised Collaborative Filtering by Text-enhanced Domain Adaptation 【阿裡】

論文：arxiv.org/abs/2007.0708

AutoML for Recommendation

1. Neural Input Search for Large Scale Recommendation Models 【Google】

論文：arxiv.org/abs/1907.0447

2. Towards Automated Neural Interaction Discovering for Click-Through Rate Prediction 【Facebook】

論文：arxiv.org/abs/2007.0643

Federated Learning

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

Evaluation

1. Counterfactual Evaluation of Slate Recommendations with Sequential Reward Interactions 【Netflix, Spotify】

論文：arxiv.org/abs/2007.1298

2. Evaluating Conversational Recommender Systems via User Simulation

論文：arxiv.org/abs/2006.0873

3. 【Best Paper Award】On Sampled Metrics for Item Recommendation 【Google】

4. On Sampling Top-K Recommendation Evaluation

Debiasing

1. Debiasing Grid-based Product Search in E-commerce 【Etsy】

論文：public.asu.edu/~rguo12/

2. Counterfactual Evaluation of Slate Recommendations with Sequential Reward Interactions 【Netflix, Spotify】

論文：arxiv.org/abs/2007.1298

3. Attribute-based Propensity for Unbiased Learning in Recommender Systems: Algorithm and Case Studies 【Google】

論文：research.google/pubs/pu

POI Recommendation

1. Geography-Aware Sequential Location Recommendation 【Microsoft】

論文：staff.ustc.edu.cn/~lian

Cold-Start Recommendation

1. MAMO: Memory-Augmented Meta-Optimization for Cold-start Recommendation

論文：arxiv.org/abs/2007.0318

2. Meta-learning on Heterogeneous Information Networks for Cold-start Recommendation

論文：https://ink.library.smu.edu.sg/cgi/...

Others

1. Improving Recommendation Quality in Google Drive 【Google】

論文：research.google/pubs/pu

2. Temporal-Contextual Recommendation in Real-Time 【Amazon】

論文：https://assets.amazon.science/96/71/...