Technologies for node-degree based clustering of data sets
US10452717B2 · kind B2 · utility
Assignee
Inventors
Key dates
| Filing date | Sep 22, 2016 |
| Grant date | Oct 22, 2019 |
| Priority date | — |
| Expiry date | Jan 8, 2038 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG06N20/00
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
Technologies for node-degree based clustering include a computing device to construct a graph that includes multiple vertices corresponding to the data points of a data set. The computing device inserts an edge between each pair of vertices that has a corresponding similarity metric that meets a predetermined threshold similarity metric. The computing device determines a node degree for each vertex in the graph and initializes a cutoff node degree as the lowest node degree of the vertices. The computing device selects a test subset of the graph that includes vertices having a node degree less than or equal to the cutoff node degree. The computing device determines whether the test subset covers the graph and if not increases the cutoff node degree. If the test subset covers the graph, the data points corresponding to the vertices of the test subset are the representative cluster. Other embodiments are described and claimed.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.