Semi-supervised learning is a machine learning approach where only a small number of training data are labeled, while most training data are not labeled. It lies somewhere between supervised and unsupervised learning and is useful when it is difficult or costly to obtain labeled data. Some common semi-supervised learning strategies include:

1. Self-training: Simple method with much use. Label only the simplest dataset, use it to label the unlabelled dataset, and retain the most certain label as the next dataset to label. Retrain the model.

2. Co-training: Two separate models are trained using different feature sets . Each model labels the other unlabelled sample so that the other model can learn the uncertain point. It can be iteratively trained to label as much data as possible.

3. Semi-Supervised Support Vector Machines: It is a semi-supervised pairwise maximum margin approach obtained by an extended SVM. S3VM tries to distinguish not only between positive and negative samples but also between labelled and unlabelled samples.

4. Graph-Based Methods: A graph is created in which nodes represent labelled or unlabelled dataset data. The graph devices work together to send-out the label and thus start using unlabelled data or device to create essentially additional labelled data called pseudolabels.

5. Generative Models: Generative models such as GMM learning generate labels for unlabeled samples with labelled data. It assumes that the data follows a specific distribution, and the unknown label is related to the most possible generating point

6. Multiview Learning: Learning multi-views of the same dataset instead of single-view datasets for unsupervised learning, that’s particularly helpful when it is difficult to have all features explain an algorithm.

7. Pseudo-Labeling: The strong labels are assumed to be correct, and new datasets can simply be added to the label.

8. Label Propagation, and Label Spreading: Label Propagation propagates the labels from a small letter to the dataset point participant’s graph that becomes a connected graph.

9. Transductive learning: A specific case of semi-supervised learning follows semi-supervised learning’s rule. forests.