Similarity Search with the Distance Density Model

Investor logo

Warning

This publication doesn't include Faculty of Arts. It includes Faculty of Informatics. Official publication website can be found on muni.cz.
Authors

KŘENKOVÁ Markéta MÍČ Vladimír ZEZULA Pavel

Year of publication 2022
Type Article in Proceedings
Conference Similarity Search and Applications: 15th International Conference, SISAP 2022, Bologna, Italy, October 5 - October 7, 2020, Proceedings
MU Faculty or unit

Faculty of Informatics

Citation
Web https://link.springer.com/chapter/10.1007/978-3-031-17849-8_10
Doi http://dx.doi.org/10.1007/978-3-031-17849-8_10
Keywords Metric space similarity model;Perceived similarity;Data-dependent similarity;Distance density model;Effective and efficient similarity search
Description The metric space model of similarity has become a standard formal paradigm of generic similarity search engine implementations. However, the constraints of identity and symmetry prevent from expressing the subjectivity and dependence on the context perceived by humans. In this paper, we study the suitability of the Distance density model of similarity for searching. First, we use the Local Outlier Factor (LOF) to estimate a data density in search collections and evaluate plenty of queries using the standard geometric model and its extension respecting the densities. We let 200 people assess the search effectiveness of the two alternatives using the web interface. Encouraged by the positive effects of the Distance density model, we propose an alternative way to estimate the data densities to avoid the quadratic LOF computation complexity with respect to the dataset size. The sketches with unbalanced bits are clarified to be in correlation with LOFs, which opens a possibility for an efficient implementation of large-scale similarity search systems based on the Distance density model.
Related projects:

You are running an old browser version. We recommend updating your browser to its latest version.