Fig. 1
From: Visual analytics for collaborative human-machine confidence in human-centric active learning tasks
Confidence-based visual analytic loop for active machine learning. User actions are shown in blue, and machine actions are shown in orange. The outer stages show the initial phase of operation, then the inner loop shows the subsequent actions for each iteration of training. Coloured arrows correspond to 4 possible forms of confidence that can be utilised from the current classifier model: selection based on confidence, and reporting of confidence, for both user and machine. Central to the process, as the user queries or selects further samples, and also confirms or refines machine-generated labels, this contributes towards their knowledge generation of the system process