Notes
0%

论文笔记-Deep Learning Based Recommender System

发表于 更新于 分类于 Valine：
该论文对当前基于深度学习的推荐系统方法做了总结和分类，并给出自己的看法

Overview

该论文发表于ACM2017， 对当前基于深度学习的推荐系统方法做了总结和分类，并给出自己的看法。论文见: Deep Learning based Recommender System： A Survey and New Perspectives

产生推荐列表的数据来源一般有如下几种: - user preferences - item features - user-item past interactions - some other features (temporal, spatial)

推荐模型主要分为三个大类: - collaborative filtering - content-based recommender system - hybrid recommender system

根据完成的任务可以分为 - rating - ranking - classification

Models

Traditional Models

  • matrix factorization (MF)
  • factorization machines (FM)
  • BPR
  • Collaborative Metric Learning (CML)
  • funkSVD

Deep Models:

deep model

其中autoencodes包含如下几种： - denoising autoencoder - marginalized denoising autoencoder - sparse autoencoder - contractive autoencoder - variational autoencoder (VAE)

  • MLP
    • Nerual Network Matrix Factorization (NNMF)
    • Nerual Collaborative Filtering (NCF)
    • Deep Factorization Machine (DeepFM)
    • Wide & deep Model
    • Deep Structured Semantic Model (DSSM)
    • Deep Semantic Similarity based Personalized Recommendation (DSPR)
    • Multi-View Deep Neural Network ( MV-DNN)
  • Autoencoder (AE)
    • AutoRec
    • CNF
    • Collaborative Denoising Auto-Encoder (CDAE)
    • Stacked denoising autoencoder (SDAE)
    • Collaborative Deep Learning (CDL)
    • Relational stacked denoising autoencoders (RSDAE)
    • Collaborative variational autoencoder (CVAE)
    • Collaborative Deep Ranking (CDR)
    • AutoSVD++
    • HDCR
  • CNN
    • Point-of-Interest (POI) recommendation
    • VPOI
    • VBPR
    • DeepCoNN
    • ConvMF
    • ConvNCF
  • RNN
    • Session-based Recommendation without User identifier
      • GRU4Rec
    • Sequential Recommendation with User Identifier
      • Recurrent Recommender Network (RRN)
  • RBM
    • RMB-CF
  • Neural Autoregressive Distribution Estimation (NADE)
    • CF-NADE
  • Adversarial Networks (AＮ)
    • IRGAN
  • Attentional Models (AM)
  • Deep Reinforcement Learning (DRL)
    • DEERS
    • DeepPage
    • DRN
  • Hybrid Models
    • CNN + AE
      • Collaborative Knowledge Based Embedding (CKE)
    • CNN + RNN
    • RNN + AE
      • CRAE
    • RNN + DRL

Future research directions

  • Join representation learning user and item content information
  • explainable recommendation with deep learning
  • going deeper for recommendation
  • machine reasoning for recommendation
  • cross domain recommendation with dnn
  • deep multi-task learning
  • scalability of dnn
  • better, more unified and harder evaluation

Top Conference

论文中提到Recommender Systems方向的定会大致有以下这些: - NIPS - ICML - ICLR - KDD - WWW - SIGIR - WSDM - RecSys