How do threshold and kbest differ in a self-training iteration?
Classic MLLevel: IntermediatePseudolabel selection policies
How do threshold and k_best differ in a self-training iteration?
Hints
- Absolute confidence versus relative ranking.
- A quota can force weak candidates through.
Answer
Threshold accepts examples above a confidence cutoff; k_best takes a fixed number of the most confident candidates. The latter can include low-confidence cases when the entire pool is uncertain.
Source and licence: scikit-learn 1.7.2 — semi supervised · BSD-3-Clause · answer: Quantoros editorial