A machine learning paradigm in which a model interactively queries a user, oracle, or environment to obtain labels or feedback for the most informative examples, iteratively improving performance while minimising annotation cost. Interactive learning encompasses active learning, online learning, and human-in-the-loop approaches that tighten the loop between model uncertainty and human input.
Semantic Classification
Content
Overview
Interactive learning is a machine learning paradigm where the model actively queries informative examples from a human oracle or environment, iteratively improving with minimal labelling effort. It encompasses active learning, online learning, and human-in-the-loop workflows. Key to this paradigm is uncertainty sampling — identifying the examples the model is least confident about — and using human feedback to refine performance in label-scarce domains.