Sharing hash codes for multiple purposes

Pronobis, Wiktor and Panknin, Danny and Kaul, Manohar and et al, . (2018) Sharing hash codes for multiple purposes. Japanese Journal of Statistics and Data Science, 1 (1). pp. 215-246. ISSN 2520-8756

Full text not available from this repository. (Request a copy)

Abstract

Locality sensitive hashing (LSH) is a powerful tool in data science, which enables sublinear-time approximate nearest neighbor search. A variety of hashing schemes have been proposed for different dissimilarity measures. However, hash codes significantly depend on the dissimilarity, which prohibits users from adjusting the dissimilarity at query time. In this paper, we propose multiple purpose LSH (mp-LSH) which shares the hash codes for different dissimilarities. mp-LSH supports L2, cosine, and inner product dissimilarities, and their corresponding weighted sums, where the weights can be adjusted at query time. It also allows us to modify the importance of pre-defined groups of features. Thus, mp-LSH enables us, for example, to retrieve similar items to a query with the user preference taken into account, to find a similar material to a query with some properties (stability, utility, etc.) optimized, and to turn on or off a part of multi-modal information (brightness, color, audio, text, etc.) in image/video retrieval. We theoretically and empirically analyze the performance of three variants of mp-LSH, and demonstrate their usefulness on real-world data sets.

[error in script]

IITH Creators: