Learning with Less Labeling (LwLL)
Abstract
The Learning with Less Labeling (LwLL) program is developing technology to greatly reduce the amount of labeled data required to train machine learning (ML) systems. In supervised ML, a system learns through the use of labeled training examples to recognize and categorize attributes of images, text, or speech. Humans provide these training-data examples to ML systems and, with enough labeled data, it is generally possible to build useful models. Obtaining large amounts of labeled data can be costly, particularly for national security applications. LwLL is addressing this problem by creating ML algorithms that learn and adapt more efficiently than current ML approaches, formally deriving the limits of machine learning and adaptation, and training with a combination of labeled and unlabeled data. LwLL aims to create ML systems that are easier to train for use in variable, unpredictable, real-world environments where training data is costly or sparse.
Document Details
- Document Type
- Accomplishment
- Publication Date
- Oct 01, 2024
- Source ID
- f5904aa2311c2dfe282c3ab6cab60b78