Revisiting Wedge Sampling for Budgeted Maximum Inner Product Search

Top-k maximum inner product search (MIPS) is a central task in many machine learning applications. This work extends top-k MIPS with budgeted setting, that asks for the best approximate given limited budget of computational operations. We investigate recent advanced sampling algorithms, including wedge and diamond sampling, to solve MIPS. First, we show essentially combination basic Our theoretical analysis empirical evaluation competitive (often superior) on approximating regarding both efficiency accuracy. Second, propose dWedge, very simple deterministic variant Empirically, dWedge provides significantly higher accuracy than other solvers while maintaining similar speedup.