EasyDeepLearn

What is active learning and when should you use it?

medium

Answer

  • Iteratively query most informative unlabeled examples for human labeling.
  • Strategies: (1) Uncertainty sampling (max entropy / min margin / least confidence).
  • (2) Query-by-committee (disagreement between ensemble).
  • (3) Diversity sampling (core-set) to avoid redundant queries.
  • (4) BALD (Bayesian mutual information).
  • Use when labels are expensive (medical, legal, expert domains).
  • Skip when labels are cheap (crowdsourcing) or class balance matters more than label count.
Check yourself — multiple choice
  • Random
  • Iteratively query most informative unlabeled: uncertainty / QBC disagreement / diversity core-set / BALD mutual info; use when labels expensive (medical/legal)
  • Random labels
  • Not real

Active learning: uncertainty / QBC / diversity / BALD; expensive labels.

#data-quality

Practise MLOps & Data Quality

215 interview questions in this topic.

Related questions