Weak Supervision
Definition
Weak supervision combines imperfect information sources into provisional labels for training data.
Generating probabilistic training labels from multiple incomplete or conflicting sources such as rules, heuristics, and external models.
Simply explained
Several uncertain cues help assign suitable labels to examples. These labels then serve as learning aids.
Why it matters in practice
The term helps to select, develop, evaluate AI systems in a well-founded manner and realistically assess their technical risks.
Example from business
An enterprise combines domain rules and existing audit results into provisional ticket labels.
Limits and pitfalls
- Heuristics can share common errors; Weak Supervision is a label generation method, not every weakly labeled learning.
Origin & history
Snorkel systematized Weak Supervision from 2016 with labeling functions and a model of their dependencies.
Use in mid-sized companies
- Possible application: An enterprise combines domain rules and existing audit results into provisional ticket labels.
Frequently asked questions
What does Weak Supervision mean?
Generating probabilistic training labels from multiple incomplete or conflicting sources such as rules, heuristics, and external models. What should you watch out for? Heuristics can share common errors; Weak Supervision is a label generation method, not every weakly labeled learning.
Sources
- Snorkel · Ratner et al., Snorkel
Source checked on 10.10.2026. Editorial application examples are hypothetical; no evidence of real customer results.
Editorially reviewed on: