Budget constrained interactive search for multiple targets

Interactive graph search leverages human intelligence to categorize target labels in a hierarchy, which is useful for image classification, product categorization, and database search. However, many existing interactive studies aim at identifying single optimally, suffer from the limitations of asking too questions not being able handle multiple targets. To address these two limitations, this paper, we study new problem <u>b</u>udget constrained <u>i</u>nteractive <u>g</u>raph <u>s</u>earch <u>m</u>ultiple targets called kBM-IGS problem. Specifically, given set T hierarchy parameters k b , goal identify -sized selections S such that closeness between as small possible, by most budget questions. We theoretically analyze updating rules design penalty function capture tackle problem, develop novel framework ask using best vertex with largest expected gain, provides balanced trade-off probability benefit gain. Based on framework, first propose an efficient algorithm STBIS SingleTarget special case kBM-IGS. Then, dynamic programming based method kBM-DP MultipleTargets further improve efficiency, heuristic but algorithms, kBM-Topk kBM-DP+. Experiments large real-world datasets ground-truths verify both effectiveness efficiency our algorithms.