/DAGNN

The DGL implementation of DAGNN

Primary LanguagePython

DAGNN

Performance

On Cora, Citeseer and Pubmed

Accuracy
Dataset Cora Citeseer Pubmed
DAGNN 84.4 ± 0.5 73.3 ± 0.6 80.5 ± 0.5
DGL 84.3 ± 0.6 73.5 ± 0.8 80.5 ± 0.3

On Arxiv

Accuracy
Dataset Arxiv
DAGNN 72.09 ± 0.25
DGL 72.39 ± 0.23

On Gowalla, Yelp2018 and Amanzon-Book

Gowalla, Yelp2018 and Amanzon-Book are datasets in the filed of recommendation systems, which are introduced in paper Neural Graph Collaborative Filtering

Recall
Dataset Gowalla Yelp2018 Amanzon-Book
NGCF 0.1556 0.0543 0.0313
DAGNN(DGL) 0.1792 0.0598 0.0408
NDCG
Dataset Gowalla Yelp2018 Amanzon-Book
NGCF 0.1327 0.0477 0.0263
DAGNN(DGL) 0.1507 0.0493 0.0315