AIExplainer
Machine Learning Intermediate 2 min read

What is self-training?

A machine learning approach where a model is trained on its own predictions or outputs to improve performance

Self-training is a technique used in machine learning where a model is trained on a small amount of labeled data and then uses its own predictions to generate more training data. This process is repeated, with the model learning from its own mistakes and improving its performance over time.

Self-training is like a student who teaches themselves a new language by speaking it with a language exchange partner. The student makes mistakes, but with each conversation, they learn from their errors and improve their language skills.

Self-training can be used to improve the accuracy of a facial recognition system. The system can be trained on a small set of labeled images and then use its own predictions to generate more training data, such as predicting the identity of individuals in unlabeled images.

Self-training is used in applications such as image classification, natural language processing, and speech recognition, where large amounts of labeled data are not available or are difficult to obtain.

One common misconception about self-training is that it is a replacement for supervised learning. However, self-training is typically used in conjunction with supervised learning, where a small amount of labeled data is used to initialize the model and then self-training is used to improve its performance.

Self-training has its roots in the 1990s, when researchers began exploring ways to improve the performance of machine learning models using unlabeled data. Since then, self-training has become a popular technique in the field of machine learning.

self-supervised learning pseudo-labeling self-labeling

Three products for different needs — explore what’s relevant to you.