Semi-supervised learning uses a small set of labelled examples together with a larger set of unlabelled data.
What the base framework establishes
The approach sits between supervised and unsupervised learning. A little supervision anchors the model, and the unlabelled data supplies scale. The framework names it as its own approach because the evidence question changes: the labelled set is too small to carry the whole evaluation, so the unlabelled set and the assumptions used to exploit it become part of the record. The clause text is not in the public sample, and this page states the approach's logic rather than the clause's wording.
Why it matters to an AI management system
Semi-supervised projects often look cheap and then fail at the evaluation stage. The governance question is whether the unlabelled data is representative and whether the small labelled set was chosen in a way that biases the result. That is a record question, not a modelling preference.
The public draft summary names no change to 7.4 Semi-supervised machine learning. The final text could still touch it.
Evidence a reviewer samples
| Category | Artefact |
|---|---|
| Label budget note | How many labels were used and how they were selected. |
| Unlabelled data statement | The source, period and coverage of the unlabelled set. |
| Evaluation design | How the small labelled set was protected for testing. |
One question to prepare for. What does the unlabelled data add, and is it from the same population?
Common gaps
- The unlabelled data comes from a different period or population than the labelled set.
- All labels are consumed in training, leaving no independent test set.
- The selection rule for labelled examples is not recorded.
Prepare before publication
Record the label budget and the selection rule, and keep a held-out slice of the labelled set for evaluation.
Dates that matter
No separate date attaches to 7.4; the record dates are the stage date, the planning estimate and the CEN expectation stated above.
Cross-references
7.2 Supervised machine learning · 7.3 Unsupervised machine learning · 8.3 Data preparation · 8.5 Verification and validation
Source
Source for 7.4: ISO/IEC 23053:2022, clause 7.4. The public sample stops before 7.4, so this page relies on the public contents and the public summary and does not quote the clause. The two official records are linked on this page so any line can be checked.
Not legal advice; does not certify. A reading of the framework and the evidence a reviewer commonly samples, not a substitute for the standard or your auditor's rules.