1. A machine learning technique where a model learns from unlabeled data by generating labels from the data itself, without requiring manual annotation.
Self-supervised training allows neural networks to learn useful representations from large amounts of unlabeled data.
O treinamento auto-supervisionado permite que redes neurais aprendam representações úteis a partir de grandes quantidades de dados não rotulados.
2. A training approach where the supervisory signal is automatically derived from the input data structure itself.
BERT uses self-supervised training to predict masked words in sentences.
BERT usa treinamento auto-supervisionado para prever palavras mascaradas em sentenças.