Table 4 Average metrics with standard deviations across 100 iterations and 5 “first-order” sets. The NDCG at rank \(n_i\) is averaged across users. S indicates an instance of SELT
From: Joint embedding–classifier learning for interpretable collaborative filtering
AUC | NS-AUC | NDCG | ||
|---|---|---|---|---|
50% | FM2 | 0.99± 0.0 | 0.92± 0.0 | 0.97± 0.0 |
CrossFM2 | 0.99± 0.0 | 0.93± 0.0 | 1.00± 0.0 | |
S-PCAf | 0.95± 0.0 | 0.70± 0.1 | 0.58± 0.2 | |
S-PCAiu | 0.95± 0.0 | 0.61± 0.2 | 0.45± 0.2 | |
S-KGE | 0.91± 0.0 | 0.43± 0.2 | 0.25± 0.2 | |
JELI | 0.99± 0.0 | 0.92± 0.1 | 0.96± 0.0 | |
65% | FM2 | 0.98± 0.0 | 0.91± 0.0 | 0.87± 0.1 |
CrossFM2 | 0.99± 0.0 | 0.91± 0.0 | 0.95± 0.0 | |
S-PCAf | 0.95± 0.0 | 0.73± 0.1 | 0.54± 0.2 | |
S-PCAiu | 0.94± 0.0 | 0.62± 0.0 | 0.34± 0.1 | |
S-KGE | 0.90± 0.0 | 0.43± 0.0 | 0.06± 0.0 | |
JELI | 0.99± 0.0 | 0.94± 0.0 | 0.94± 0.1 | |
80% | FM2 | 0.97± 0.0 | 0.84± 0.1 | 0.56± 0.1 |
CrossFM2 | 0.98± 0.0 | 0.87± 0.0 | 0.74± 0.0 | |
S-PCAf | 0.95± 0.0 | 0.73± 0.1 | 0.38± 0.1 | |
S-PCAiu | 0.93± 0.0 | 0.62± 0.1 | 0.20± 0.0 | |
S-KGE | 0.91± 0.0 | 0.55± 0.1 | 0.12± 0.1 | |
JELI | 0.99± 0.0 | 0.94± 0.0 | 0.85± 0.2 |