Active Learning
A specialized training paradigm where the model explicitly selects the most informative unlabelled data points and queries a human annotator for labels.
Think of It Like This
Like a student who only asks the teacher for help on the hardest questions they don't understand, rather than asking about everything in the textbook.
Instead of randomly annotating huge datasets, active learning relies on uncertainty sampling to find the exact examples that will maximize performance gains. This drastically reduces the total cost and time required for data labelling by ignoring redundant or easy-to-classify samples. It is especially critical in domains like medical imaging where expert annotation time is prohibitively expensive.