Active Learning refers to an iterative process which consists of adding new examples to a training dataset with the goal of improving model performance, specially for certain subsets of the data. For example, if an image recognition model is having a hard time differentiating zebras from horses, we can add more samples from those categories to the training set.