Table 3:
Experiments results of using features and not using features for both CAML and HCAML under their best hyper-parameters
MODEL | HIT@1 | HIT@5 | AUC |
---|---|---|---|
CAML | 0.6611 | 0.8432 | 0.9444 |
CAML+Features | 0.687 | 0.8818 | 0.9376 |
HCAML | 0.6974 | 0.8918 | 0.9514 |
HCAML+Features | 0.6969 | 0.8903 | 0.9496 |