Fast Cross-polytope Locality-sensitive Hashing

Kennedy Christopher, Ward Rachel. Arxiv 2016

[Paper]    
ARXIV Independent LSH

We provide a variant of cross-polytope locality sensitive hashing with respect to angular distance which is provably optimal in asymptotic sensitivity and enjoys $\mathcal{O}(d\ln d)$ hash computation time. Building on a recent result (by Andoni, Indyk, Laarhoven, Razenshteyn, Schmidt, 2015), we show that optimal asymptotic sensitivity for cross-polytope LSH is retained even when the dense Gaussian matrix is replaced by a fast Johnson-Lindenstrauss transform followed by discrete pseudo-rotation, reducing the hash computation time from $\mathcal{O}(d^2)$ to $\mathcal{O}(d\ln d)$. Moreover, our scheme achieves the optimal rate of convergence for sensitivity. By incorporating a low-randomness Johnson-Lindenstrauss transform, our scheme can be modified to require only $\mathcal{O}({\ln }^9(d))$ random bits

Similar Work

• Practical And Optimal LSH For Angular Distance
• Practical And Optimal LSH For Angular Distance
• Lattice-based Locality Sensitive Hashing Is Optimal
• Optimal Data-dependent Hashing For Approximate Near Neighbors