Figure 2

Random Learning vs Active Learning Flowchart. Comparison of Random Learning (RL, top row) and Active Learning (AL, bottom row) training processes. In RL, unlabeled data (a) is sent to an expert (b), who assigns a label to each sample in the image (c): red regions indicate cancer, and green indicates non-cancer. These labeled samples are used to train a supervised classifier (d). In AL, unlabeled samples (e) are analyzed to find informative samples (f), and only informative samples (g) are annotated for training (h). The supervised classifier (i) can be re-trained and used to identify new samples that may be informative. In the AL setup, only new samples that will improve classification accuracy are added.