A note on the universal approximation capability of support vector machines

Autor(en): Hammer, B
Gersmann, K
Stichwörter: approximation; classification; Computer Science; Computer Science, Artificial Intelligence; kernel; NEURAL-NETWORKS; support vector machine; universal approximation capability
Erscheinungsdatum: 2003
Herausgeber: SPRINGER
Journal: NEURAL PROCESSING LETTERS
Volumen: 17
Ausgabe: 1
Startseite: 43
Seitenende: 53
Zusammenfassung: 
The approximation capability of support vector machines (SVMs) is investigated. We show the universal approximation capability of SVMs with various kernels, including Gaussian, several dot product, or polynomial kernels, based on the universal approximation capability of their standard feedforward neural network counterparts. Moreover, it is shown that an SVM with polynomial kernel of degree p - 1 which is trained on a training set of size p can approximate the p training points up to any accuracy.
ISSN: 13704621
DOI: 10.1023/A:1022936519097

Show full item record

Google ScholarTM

Check

Altmetric