Semi-Supervised Learning
Semi-supervised learning trains a model on a small amount of labelled data combined with a much larger amount of unlabelled data.
What is Semi-Supervised Learning?
Labelling data is often slow and expensive, while unlabelled data is plentiful. Semi-supervised methods use the labelled examples to guide learning and the unlabelled examples to learn the overall shape of the data, often reaching accuracy close to fully supervised training at lower cost.
How it works
A common technique is pseudo-labelling: a model trained on the labelled data predicts labels for the unlabelled data, and its most confident predictions are added to the training set.
← All Artificial Intelligence & Superintelligence definitions · ← Back to the A–Z definitions