Method and system for incrementally learning an adaptive subspace by optimizing the maximum margin criterion
US7502495B2 · kind B2 · utility
Assignee
Inventors
Key dates
| Filing date | Mar 1, 2005 |
| Grant date | Mar 10, 2009 |
| Priority date | — |
| Expiry date | Sep 7, 2026 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG06F18/2132
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
A method and system for generating a projection matrix for projecting data from a high dimensional space to a low dimensional space. The system establishes an objective function based on a maximum margin criterion matrix. The system then provides data samples that are in the high dimensional space and have a class. For each data sample, the system incrementally derives leading eigenvectors of the maximum margin criterion matrix based on the derivation of the leading eigenvectors of the last data sample. The derived eigenvectors compose the projection matrix, which can be used to project data samples in a high dimensional space into a low dimensional space.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.