Kernel ridge regression model for sediment transport in open channel flow

Sediment transport modeling is of primary importance for the determination channel design velocity in lined channels. This study proposes to model sediment open flow using kernel ridge regression (KRR), a nonlinear technique formulated reproducing Hilbert space. While naïve approach provides high flexibility purposes, regularized variant equipped with an additional mechanism better generalization capability. In order tailor KRR problem, unlike conventional approach, this parameter directly learned from data via new gradient descent-based learning mechanism. Moreover, construction, procedure based on Cholesky decomposition and forward-back substitution applied improve computational complexity approach. Evaluation recommended performed utilizing large number laboratory experimental where examination proposed terms three statistical performance indices indicates developed particle Froude computation, outperforming models as well some other machine techniques.