“…Recently there are modified PLS methods for SAR modeling [Bennett et al, 2003;Rosipal and Krämer, 2006b], in which kernel partial least squares (kernel PLS) is a kernel method for PLS that introduces on-linear mapping through kernels [Lapinsh et al, 2005;Deng et al, 2004]. Many other learning algorithms are applcable for SAR modeling and include Neural Networks (NN) [Tetko et al, 2001;Livingstone and Manallack, 2003;Guha and Jurs, 2005], Decision Tree [Sussman et al, 2003], Recursive Partitioning [Chen et al, 1998;Rusinko et al, 1999;An and Wang, 2001], Linear Discriminant Analysis (LDA) [Otto, 1999], Bayesian models [Xia et al, 2004;Mccallum and Nigam, 1998], and Random Forest [Zhang and Aires-de Sousa, 2007;Breiman, 2001]. Hughes-Oliver et al [2008] implement and compare many popular learning methods for SAR modeling.…”