Figure 1.

Supervised or unsupervised training. (A) Supervised training which can predict classification or regression of the input data. (B) Unsupervised training which can cluster the data. (C) Semi-supervised training. First, train the algorithm by small number of labeled data. Then, use trained algorithm to “label” unlabeled data. Next, re-train algorithm using newly labeled data and originally labeled data. Finally, algorithm trained with all data and can predict classification or regression of the input data.