Name
Abstract | ||
|---|---|---|
Traditional methods of constructing of least square support vector regression (LSSVR) do not consider the gradients of the true function but just think about the exact responses at samples. If gradient information is easy to get, it should be used to enhance the surrogate. In this paper, the gradient-enhanced least square support vector regression (GELSSVR) is developed with a direct formulation by incorporating gradient information into the traditional LSSVR. The efficiencies of this technique are compared by analytical function fitting and two real life problems (the recent U.S. actuarial life table and Borehole). The results show that GELSSVR provides more reliable prediction results than LSSVR alone. |
Year | DOI | Venue |
|---|---|---|
2016 | 10.1007/s11063-014-9402-5 | Neural Processing Letters |
Keywords | Field | DocType |
Least square support vector regression,Machine learning,Gradient information | Least squares,Data mining,Support vector machine,Analytic function,Artificial intelligence,Machine learning,Mathematics | Journal |
Volume | Issue | ISSN |
43 | 1 | 1370-4621 |
Citations | PageRank | References |
4 | 0.46 | 16 |
Authors | ||
2 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Xiaojian Zhou | 1 | 74 | 9.19 |
| Ting Jiang | 2 | 89 | 18.32 |