modAL-python/modAL

Provides pluggable uncertainty measures and query strategies—including entropy sampling, margin sampling, and custom implementations—that work seamlessly with scikit-learn and Keras models. Built around a modular architecture where the learner, estimator, uncertainty measure, and query strategy are independently swappable, enabling rapid experimentation with novel active learning algorithms without rewriting core loop logic.

2,342 stars. No commits in the last 6 months.