推薦系統：Attention Model

Attention Model in Recommender Systems: A Review

[1] Chen, Jingyuan, et al. "Attentive collaborative filtering: Multimedia recommendation with item-and component-level attention." In SIGIR, 2017.

idea: there exists item- and component-level implicitness which blurs the underlying users preferences in multimedia recommendation

model: a neural network that consists of two attention modules: the component-level attention module, starting from any content feature extraction network (e.g., CNN for images/videos), which learns to select informative components of multimedia items, and the item-level attention module, which learns to score the item preferences.

framework:

Item-level attention:

Component-Level attention:

[2] Xiao, Jun, et al. "Attentional factorization machines: Learning the weight of feature interactions via attention networks." In IJCAI, 2017

framework:

Factorization Machines:

Attentional Factorization Machines:

[3] He, Xiangnan, et al. "NAIS: Neural Attentive Item Similarity Model for Recommendation." TKDE, 2017

framework:

trick: is the smoothing exponent

[4] Li, Jing, et al. "Neural Attentive Session-based Recommendation." In CIKM, 2017.

idea: explore a hybrid encoder with an attention mechanism to model the users sequential behavior and capture the users main purpose in the current session

framework:

Global Encoder: model the users sequential behaviors

Local Encoder: model the users main purpose

fusion:

[5] Wang, Shoujin, et al. "Attention-based Transactional Context Embedding for Next-Item Recommendation." In AAAI, 2018

idea: next basket recommendation, not only consider all the observed items in the current transaction, but also weight them with different relevance to build an attentive context

framework:

[6] Wang, Xiang, et al. 「TEM Tree-enhanced Embedding Model for Explainable Recommendation.」 In WWW, 2018

idea: propose a Tree-enhanced Embedding Method that combines the strengths of embedding-based and tree-based models, first employ a tree-based model to learn explicit decision rules from the rich side information, then design an embedding model that can incorporate explicit cross features and generalize to unseen cross features on user ID and item ID

framework:

（本文持續更新）

推薦閱讀：