Excuse me, while reading your code, I found that your aux_objective_loss calculation uses all the true labels. However, your paper states that only a portion of the true labels should be used. If all the true label information is already known, why would clustering be necessary? Does this conflict with the essence of clustering tasks?
Excuse me, while reading your code, I found that your aux_objective_loss calculation uses all the true labels. However, your paper states that only a portion of the true labels should be used. If all the true label information is already known, why would clustering be necessary? Does this conflict with the essence of clustering tasks?