An improved algorithm for locality-sensitive hashing

Abstract

We present an improved Locality-Sensitive Hashing for similarity search under high dimension search. Our scheme improves the running time based on the earlier algorithm Locality-Sensitive Hashing for hamming distance and euclidean distance. In this paper we have collected a database of The MNIST DATABASE, we proposed nearest neighbor search in the database and can get a good result in an acceptable time. The experimental results show that our data structure is up to about 10 times faster than ordinary Locality-Sensitive Hashing when working on a database of 60000 samples. At the same time, the accuracy rate and recall rate are higher than earlier algorithms.