AAAI2020推薦系統論文集錦

  • 2020 年 2 月 14 日
  • 筆記

前言

最近整理了AAAI2020會議中關於推薦系統的論文,同時通過代碼分析了下所接收論文的標題,發現了一些研究的熱點以及趨勢。

概述

通過對所接收的1590篇論文的標題進行分析,發現以下結論:

  • 大部分的論文所用到的技術多為Neural Network(128)相關的;
  • 大部分的文獻聚焦在以下幾個關鍵技術。比如Embedding(51), Attention(49), Adversarial(74), Reinforcement(49), Convolutional(42), Recurrent(16)等;
  • 主要面向的研究任務有分類、回歸、識別、追蹤等,其中推薦的比重所佔也不小。比如Classification(50), Regression(15), Prediction(39), Recognition(52), Tracking(20), Segmentation(28), Translation(32), Recommendation(21)等;
  • 所研究的數據不僅關注準確性,關注指標更加多樣化。比如Efficient(59), Robust(30), Dynamic(29), Adaptive(29), Hierarchical(26),;
  • 論文研究所用到的數據以圖為主,視頻、圖像、文本比重相當。比如Graph(128), Video(35), Image(59), Heterogeneous(16), Text(35), Social(20)。

*其中括號里的數字表示出現次數。

推薦相關的文章

特此從1590篇論文中篩選出與推薦相關的27篇文章供大家提前閱讀,提前領略牛人的最新想法。

  • PEIA: Personality and Emotion Integrated Attentive Model for Music Recommendation on Social Media.
  • Where to Go Next: Modeling Long-and Short­‐Term User Preferences for Point-­of‐Interest Recommendation.
  • A Knowledge-­Aware Attentional Reasoning Network for Recommendation.
  • Enhancing Personalized Trip Recommendation with Attractive Routes.
  • Efficient Heterogeneous Collaborative Filtering without Negative Sampling for Recommendation.
  • An Attentional Recurrent Neural Network for Personalized Next Location Recommendation.
  • Memory Augmented Graph Neural Networks for Sequential Recommendation.
  • Leveraging Title-Abstract Attentive Semantics for Paper Recommendation.
  • Diversified Interactive Recommendation with Implicit Feedback.
  • Question-­driven Purchasing Propensity Analysis for Recommendation.
  • Sequential Recommendation with Relation-­Aware Kernelized Self-­Attention.
  • Incremental Fairness in Two­‐Sided Market Platforms: On Smoothly Updating Recommendations.
  • Attention‐guide Walk Model in Heterogeneous Information Network for Multi-­style Recommendation.
  • Table2Analysis: Modeling and Recommendation of Common Analysis Patterns for Multi-­Dimensional Data.
  • Symmetric Metric Learning with Adaptive Margin for Recommendation.
  • Multi-­Feature Discrete Collaborative Filtering for Fast Cold-­start Recommendation.
  • Towards Comprehensive Recommender Systems: Time-­Aware Unified Recommendations Based on Listwise Ranking of Implicit Cross-Network Data.
  • Fast Adaptively Weighted Matrix Factorization for Recommendation with Implicit Feedback.
  • Towards Hands‐free Visual Dialog Interactive Recommendation.
  • Contextual-­Bandit Based Personalized Recommendation with Time-­Varying User Interests.
  • Stochastically Robust Personalized Ranking for LSH Recommendation Retrieval.
  • Revisiting Graph based Collaborative Filtering: A Linear Residual Graph Convolutional Network Approach.
  • Multi-Component Graph Convolutional Collaborative Filtering.
  • Deep Match to Rank Model for Personalized Click-Through Rate Prediction.
  • Deep Time-Stream Framework for Click-Through Rate Prediction by Tracking Interest Evolution.
  • Improved Algorithms for Conservative Exploration in Bandits.
  • Linear Bandits with Feature Feedback.

總結

隨着推薦系統的重要性越來越大,研究推薦的人逐漸在增多;隨着工業界所產生的用戶數據越來越多,工業界研究推薦的優勢也越來越大。此次會議上出現了許多推薦的應用,比如音樂推薦、興趣點推薦、旅遊推薦、論文推薦等;同時也有相關的研究放到冷啟動、推薦效率等問題上。