Machine Learning
274 views
Self-Supervised Learning
Quick Definition
Generating supervisory signals from unlabeled data for training
Full Definition
A ML approach where the model generates its own supervisory signal from unlabeled data.
Examples
BERT pre-training, contrastive learning, masked modeling
Related Terms
contrastive-learning
transfer-learning
More Machine Learning Terms
LightGBM
Fast gradient boosting framework using histogram splitting
Decision Tree
Tree-like model splitting data based on feature values
Dropout
Regularization randomly deactivating neurons during training
Gaussian Mixture Model
Probabilistic model using multiple Gaussian distributions
K-Nearest Neighbors
Classifying based on majority class of k nearest neighbors
Semi-Supervised Learning
Combining small labeled datasets with large unlabeled data