chapter  22
Active Learning: A Survey
ByCharu C. Aggarwal, Xiangnan Kong, Quanquan Gu, Jiawei Han, Philip S. Yu
Pages 36

In all these cases, labels can be obtained, but only at a significant cost to the end user. An important observation is that all records are not equally important from the perspective of labeling. For example, some records may be noisy and contain no useful features that are relevant to classification. Similarly, records that cleanly belong to one class or another may be helpful, but less so than records that lie closer to the separation boundaries between the different classes.