English

Improved approximate near neighbor search without false negatives for $l_2$

Computational Geometry 2017-10-02 v1

Abstract

We present a new algorithm for the cc--approximate nearest neighbor search without false negatives for l2dl_2^d. We enhance the dimension reduction method presented in \cite{wygos_red} and combine it with the standard results of Indyk and Motwani~\cite{motwani}. We present an efficient algorithm with Las Vegas guaranties for any c>1c>1. This improves over the previous results, which require c=ω(loglogn)c=\omega(\log\log{n}) \cite{wygos_red}, where nn is the number of the input points. Moreover, we improve both the query time and the pre-processing time. Our algorithm is tunable, which allows for different compromises between the query and the pre-processing times. In order to illustrate this flexibility, we present two variants of the algorithm. The "efficient query" variant involves the query time of O(d2)O(d^2) and the polynomial pre-processing time. The "efficient pre-processing" variant involves the pre-processing time equal to O(dω1n)O(d^{\omega-1} n) and the query time sub-linear in nn, where ω\omega is the exponent in the complexity of the fast matrix multiplication. In addition, we introduce batch versions of the mentioned algorithms, where the queries come in batches of size dd. In this case, the amortized query time of the "efficient query" algorithm is reduced to O(dω1)O(d^{\omega -1}).

Keywords

Cite

@article{arxiv.1709.10338,
  title  = {Improved approximate near neighbor search without false negatives for $l_2$},
  author = {Piotr Wygocki},
  journal= {arXiv preprint arXiv:1709.10338},
  year   = {2017}
}

Comments

arXiv admin note: text overlap with arXiv:1708.06395