An Indexing Structure for Dynamic Multidimensional Data in Vector Space by Elena Mikhaylova, Boris Novikov and Anton Volokhov. (Advances in Databases and Information Systems, Advances in Intelligent Systems and Computing, 2013, Volume 186, 185-193, DOI: 10.1007/978-3-642-32741-4_17)
Abstract:
The multidimensional k – NN (k nearest neighbors) query problem is relevant to a large variety of database applications, including information retrieval, natural language processing, and data mining. To solve it efficiently, the database needs an indexing structure that provides this kind of search. However, attempts to find an exact solution are hardly feasible in multidimensional space. In this paper, a novel indexing technique for the approximate solution of k – NN problem is described and analyzed. The construction of the indexing tree is based on clustering. Indexing structure is implemented on top of high-performance industrial DBMS.
The review of recent work is helpful but when the paper reaches the algorithm for indexing “…dynamic multidimensional data…,” it slips away from me.
Where is the dynamic nature of the data that is being overcome by the indexing?
I ask because we are human observers are untroubled by the curse of dimensionality, even when data is dynamically changing.
Although those are two important aspects when we process it by machine:
- The number of dimensions of data, and
- The rate at which the data is changing.