Application of the nonlinear dimensionality reduction method with the use of reference nodes for solving the problem of multidimensional data search

Abstract

This paper considers a two-step approach to solving the problem of multidimensional data search in which, first, the dimensionality of the original data set is reduced and only then the index structure is built in the low-dimensional space. To reduce the dimensionality, the nonlinear dimensionality reduction method with the use of reference nodes is applied. R-trees are used as the index structure. The carried out experiments show that this approach can be used for rapid prefiltering of multidimensional data.