A Bandit Technique to Most Inside Product Search
Authors: Rui Liu, Tianyi Wu, Barzan Mozafari
Abstract: There was substantial evaluation on sub-linear time approximate algorithms for Most Inside Product Search (MIPS). To achieve fast query time, state-of-the-art methods require essential preprocessing, which might be a burden when the number of subsequent queries simply is not sufficiently large to amortize the payment. Furthermore, present methods mustn’t have the pliability to straight administration the suboptimality of their approximate outcomes with theoretical ensures. On this paper, we advise the first approximate algorithm for MIPS that does not require any preprocessing, and permits clients to handle and positive the suboptimality of the outcomes. We stable MIPS as a Best Arm Identification draw back, and introduce a model new bandit setting which will completely exploit the actual building of MIPS. Our technique outperforms state-of-the-art methods on every synthetic and real-world datasets. △
