A machine learning approach that combines small amounts of labeled data with larger amounts of unlabeled data during training. This technique leverages the abundance of unlabeled data to improve model performance when labeled examples are scarce or expensive.
