Reviews: Efficient nonmyopic batch active search

This work investigates the different batch-mode extensions of an active search method called efficient nonmyopic policy (ENS) [12]. ENS achieves good performance efficiently because it assumes the sample selections are independent after a step [12]. This paper proposes two strategies: 1) converting the batch active search problem to sequential one by guessing the hidden labels of selected samples 2) try to enumerate all possible hidden labels of selected samples by Monte Carlo. Strength: Allowing to sample batches is important in practice. The work addresses several theoretical and practical challenges in many aspects of batch active search such as how difficult batch active search could be, why pessimistic oracle works well, and how to make the methods more efficient by pruning.