1. Active Query Driven by Uncertainty and Diversity for Incremental Multi-Label Learning
The key task in active learning is to design a selection criterion such that queried labels can improve the classification model most.
many active selection criteria:
uncertainty measures the confidence of the current model on classifying an instance ,
diversity measures how different an instance is from the labeled data ,
density measures the representativeness of an instance to the whole data set .
In traditional supervised classification problems, one instance is assumed to be associated with only one label. However, in many real world applications, an object can have multiple labels simultaneously. Multi-label learning is a framework dealing with such objects.