100 Facts About Semi-supervised Learning

Photo by Donny Jiang on Unsplash

Here are 100 technical facts about semi-supervised learning:

1. Definition: Semi-supervised learning is a machine learning paradigm where a model is trained on a dataset containing both labeled and unlabeled data.
2. Labeling Ratio: In semi-supervised learning, the proportion of labeled to unlabeled data is typically imbalanced.
3. Advantage over Supervised Learning: Semi-supervised learning can leverage large amounts of unlabeled data, often outperforming purely supervised models.
4. Transductive Learning: Semi-supervised learning can be transductive, where the model makes predictions for specific unlabeled instances in the training set.
5. Inductive Learning: In semi-supervised learning, inductive approaches involve learning a general model from the labeled data and applying it to all instances, labeled and unlabeled.
6. Self-Training: A common semi-supervised technique where a model iteratively trains on the most confidently predicted unlabeled samples.
7. Co-Training: Using multiple models trained on different feature subsets or views of the data and exchanging information.
8. Multi-View Learning: Learning from multiple representations of the data, potentially reducing the impact of noise.