I have a multi-label classification problem, where each data point has exactly 3 labels (out of many labels, say 1000). In my model, I pick the top 5 predicted labels.
He