Paper
[WWW2021-FairGo] Learning Fair Representations for Recommendation: A Graph-based Perspective [pdf] [code]
One word: The correlation between users and items is not considered in fair recommendation model. Removing the sensitive attributes of the user and user-centric graph embedding, then achieving fairness.
[AAAI2020-LR-GCCF] Revisiting Graph based Collaborative Filtering: A Linear Residual Graph Convolutional Network Approach [pdf] [code]
One word: Focusing on over-smoothing problem and training difficulty problem. Removing non-linear activations and designing residual network structure to solve problem.
[TKDE2019-HASC] A Hierarchical Attention Model for Social Contextual Image Recommendation [pdf arxiv] [code]
One word: Focuing on fuseing various kinds of side information(e.g., image visual representation, social network). We develop a hierarchical attention model for social contextual image recommendation.
Dataset
Flickr-collect and organize by myself
In TKDE2019-HASC, we collected and published Flickr data with rich information.
-
Favor <user,item> matrix, e.g.:<1,2> denotes that user(id:1) favors item(id:2)
-
Upload History <user, upload history item> matrix, e.g.:<1,[7,11,23]> show that user(id:1) uploads some items(id:7,11 and 23)
-
Social Neighborhood <user, follow user> matrix, e.g.:<1,[5,6]> show that user(id:1) follows some users(id:5,6)
-
Owner Admiration <item, owner user> matrix, e.g:<7,1> show that items(id:7) is uploaded by user(id:1)
-
Item Visual Information <item, image visual feature> matrix, e.g:<2,[]4096> is the image feature of item(id:2)
F_L dataset, ensure each user and each image have at least 10 rating records. [download_link]
F_S dataset, a smaller but denser dataset. [download_link]