Improved Signi¯cant Vector Learning forParsimonious Regression
-
Graphical Abstract
-
Abstract
Although the Signi¯cant vector (SV) re-
gression algorithm was proposed for constructing parsimo-
nious regression model, yet it can only ¯nd a suboptimal
solution. Hence, in this paper, an improved scheme is pro-
posed to boost the performance of the SV algorithm. The
Improved signi¯cant vector (ISV) algorithm without the
complicated regularization technique achieves as excellent
results as the Regularized SV algorithm (RSV), meantime
with less number of regressors. In addition, compared
with other algorithms, i.e., local regularization assisted or-
thogonal least squares (LROLS), Relevance vector machine
(RVM), and the Modi¯ed Gram-Schmidt (MGS), the ISV
algorithm is also favorable in the parsimoniousness. Fi-
nally, simulation examples validate the e®ectiveness and
feasibility of the ISV algorithm.
-
-
DownLoad: