요약: batch 단위로 어노테이터에게 쿼리를 보내는 미니 배치 기반으로 Diversity와 Informativeness를 결합한 AL 방법론이다. K-means를 사용하여 iteration 별 k개의 센터와 가장 가까운 인스턴스들을 쿼리하여 Diversity를 기반으로 데이터를 쿼리하는데, 이 때 K-means의 목적함수에 Informativeness score(Marginal Sampling 방식으로 Uncertainty를 측정한 스코어)를 추가하여 Informativeness 까지 고려하였다.