Title :
Training in pattern recognition from a small number of observations using projections onto null-space
Author :
Fursov, Vladimir A.
Author_Institution :
Image Processing Syst. Inst., Acad. of Sci., Samara, Russia
Abstract :
Based on the algebraic approach, a system is trained in pattern recognition from a small number of observations. The feature space is built and tolerances are set up using a null-space of matrices composed of the feature vectors of the learning objects
Keywords :
eigenvalues and eigenfunctions; feature extraction; learning (artificial intelligence); matrix algebra; pattern recognition; eigenvectors; feature vectors; learning; matrix algebra; null-space; pattern recognition; Character generation; Computational efficiency; Eigenvalues and eigenfunctions; Image processing; Laboratories; Pattern recognition; Sampling methods; Vectors; Yttrium;
Conference_Titel :
Pattern Recognition, 2000. Proceedings. 15th International Conference on
Conference_Location :
Barcelona
Print_ISBN :
0-7695-0750-6
DOI :
10.1109/ICPR.2000.906192
