CTR prediction on Last.FM (test)
80.3AUCKGNN-LS
Evaluation results
| Method | Links | ||
|---|---|---|---|
| KGNN-LS2019.05 | 80.3 | — | |
| MKR2019.01 | 79.7 | 75.2 | |
| MKR-1LHigh-level layers K=12019.01 | 79.5 | 74.9 | |
| MKR-DCNArchitecture=DCN-based variant2019.01 | 77.8 | 73 | |
| LibFMdimension={1, 1, 8}, training epochs=502019.05 | 77.8 | — | |
| LibFM2019.01 | 77.7 | 70.9 | |
| LibFM + TransETransE dimension=322019.05 | 77.7 | — | |
| MKR-stitchArchitecture=cross-stitch network variant2019.01 | 77.2 | 72.5 | |
| RippleNetd=16, H=3, lambda1=10^-5, lambda2=0.02, eta=0.0052019.05 | 77 | — | |
| SVD2019.05 | 76.9 | — | |
| RippleNet2019.01 | 76.8 | 69.1 | |
| Wide&Deep2019.01 | 75.6 | 68.8 | |
| CKE2019.01 | 74.4 | 67.3 | |
| CKEembedding dimension=64, KG training weight=0.12019.05 | 74.4 | — | |
| PER2019.01 | 63.3 | 59.6 | |
| PERmeta-paths=user-musician-date_of_birth-musician, user-musician-country-musician, user-musician-genre-musician2019.05 | 63.3 | — | |
| DKN2019.01 | 60.2 | 58.1 |