Transductive Learning
Direct learningIt is a method of predicting specific test samples by observing specific training samples.
Direct learning, like semi-supervised learning, does not require human intervention. The difference is that direct learning assumes that the unlabeled data is the data used for testing, and its purpose is to achieve the best generalization ability on this data; while semi-supervised learning does not know what the final test case will be during learning.
In other words, direct learning is similar to a subproblem of semi-supervised learning, or a specialized form of semi-supervised learning, so it is sometimes also classified as semi-supervised learning.
Active learning, pure semi-supervised learning, and direct learning are all learning techniques that use unlabeled data, but they have slight differences in ideas.