Fig. 8
From: Visual analytics for collaborative human-machine confidence in human-centric active learning tasks
Example to show limitations of distance-based selection scheme. 60 samples have been selected by the machine and a cluster of 8’s can be observed. Yet, there is a 3 directly by a selected sample in this cluster (manually selected by user). This mis-classification would not be detected by distance-selection until the distance parameter is sufficiently low for a neighbouring point to be selected. A collaborative approach that incorporates other machine-selection schemes such as machine-confidence random selection (MCRS) scheme would help to overcome these limitations