In the case of the music dataset AUC, the original was close to 80%, but this version gets to 85%,
In the case of the music dataset AUC, the original was close to 80%, but this version gets to 85%,