Fast Similarity Search Using Multiple Binary Codes

Abstract

One of the fast similarity search techniques is a binary hashing method that transforms a real-valued vector into a binary code. The similarity between two binary codes is measured by their Hamming distance. In this method, a hash table is often used for realizing the constant time similarity search. The number of accesses to the hash table, however, increases when the number of bits becomes long. In this paper, we consider the method that does not access the data with long Hamming radius by using multiple binary codes. Then, we propose the learning method of the binary hash functions for multiple binary codes. We conduct the experiment on similarity search utilizing up to 20 million data set, and show that our proposed method achieves a fast similarity search compared with the conventional linear scan and hash table search.