A survey on instance selection for active learning

Active learning aims to train an accurate prediction model with minimum costby labeling most informative instances. In this paper, we survey existing works on activelearning from an instance-selection perspective and classify them into two categories with aprogressive relationship: (1) active learning merely based on uncertainty of independent andidentically distributed (IID) instances, and (2) active learning by further taking into accountinstance correlations. Using the above categorization, we summarize major approaches inthe field, along with their technical strengths/weaknesses, followed by a simple runtimeperformance comparison, and discussion about emerging active learning applications andinstance-selection challenges therein. This survey intends to provide a high-level summarization for active learning and motivates interested readers to consider instance-selectionapproaches for designing effective active learning solutions.