active_learning_steps

probly.evaluation.active_learning.loop.active_learning_steps(pool: ActiveLearningPool, estimator: E, query_strategy: QueryStrategy[E], query_size: int = 1000, n_iterations: int = 10) → Iterator[ALState[E]]

Yield AL state after initial training and each query-retrain cycle.

Parameters:

• pool – Data pool managing labeled/unlabeled splits.

• estimator – Model to train and query. Must implement fit/predict/predict_proba.

• query_strategy – Strategy for selecting which unlabeled samples to query.

• query_size – Number of samples to query per iteration.

• n_iterations – Maximum number of query-retrain iterations.

Yields:

ALState after initial training (iteration=0) and after each subsequent query-retrain cycle.