Abstract
Original language | English |
---|---|
Title of host publication | 17th International Conference on Pattern Recognition (ICPR'04) |
Place of Publication | USA |
Publisher | IEEE |
Pages | 359-362 |
Number of pages | 4 |
Volume | 3 |
ISBN (Electronic) | 0769521282 |
Publication status | Published - 2004 |
Event | International Conference on Pattern Recognition - Cambridge UK, United Kingdom Duration: 23 Aug 2004 → 26 Aug 2004 |
Conference
Conference | International Conference on Pattern Recognition |
---|---|
Country | United Kingdom |
Period | 23/08/04 → 26/08/04 |
Fingerprint
Cite this
}
Critical Vector Learning to Construct RBF Classifiers. / Shi, Daming; Ng, G.S.; Gao, Junbin; Yeung, D.S.
17th International Conference on Pattern Recognition (ICPR'04). Vol. 3 USA : IEEE, 2004. p. 359-362.Research output: Book chapter/Published conference paper › Conference paper
TY - GEN
T1 - Critical Vector Learning to Construct RBF Classifiers
AU - Shi, Daming
AU - Ng, G.S.
AU - Gao, Junbin
AU - Yeung, D.S.
N1 - Imported on 03 May 2017 - DigiTool details were: publisher = USA: IEEE, 2004. Event dates (773o) = 23-08-2004-26-08-2004; Parent title (773t) = International Conference on Pattern Recognition.
PY - 2004
Y1 - 2004
N2 - Sensitivity is initially investigated for the construction of a network prior to its design. Sensitivity analysis applied to network pruning seems particularly useful and valuable when network training involves a large amount of redundant data. This paper proposes a novel learning algorithm for the construction of radial basis function (RBF) classifiers using sensitive vectors (SenV), to which the output is the most sensitive. In training, the number of hidden neurons and the centers of their radial basis functions are determined by the maximization of the output's sensitivity to the training data. In classification, the minimal number of such hidden neurons with the maximal sensitivity will be the most generalizable to unknown data. Our experimental results suggests that our proposed methodology outperforms classical RBF classifiers constructed by clustering.
AB - Sensitivity is initially investigated for the construction of a network prior to its design. Sensitivity analysis applied to network pruning seems particularly useful and valuable when network training involves a large amount of redundant data. This paper proposes a novel learning algorithm for the construction of radial basis function (RBF) classifiers using sensitive vectors (SenV), to which the output is the most sensitive. In training, the number of hidden neurons and the centers of their radial basis functions are determined by the maximization of the output's sensitivity to the training data. In classification, the minimal number of such hidden neurons with the maximal sensitivity will be the most generalizable to unknown data. Our experimental results suggests that our proposed methodology outperforms classical RBF classifiers constructed by clustering.
KW - Open access version available
KW - Radial basis function networks
KW - Sensitivity analysis
M3 - Conference paper
VL - 3
SP - 359
EP - 362
BT - 17th International Conference on Pattern Recognition (ICPR'04)
PB - IEEE
CY - USA
ER -