Application of Support Vector Regression in Krylov Solvers

Support vector regression (SVR) is well known as a or prediction tool under the Machine Learning (ML) which preserves all key features through training data. Different from general prediction, here, we proposed SVR to predict new approximate solutions after generated some iterates using an iterative method called Lanczos algorithm, one class of Krylov solvers. As know that solvers, including methods, for solving high dimensions systems linear equations (SLEs) problems experiences breakdown causes sequence incomplete, good solution never reached. By assuming exist breakdown, then could what they are. It realized by learning previous also The used next iterate expected now has similar property before breaking down. Furthermore, implemented hybrid SVR-Lanczos (or SVR-L) in restarting frame work, it restarting-SVR-L. idea behind time running SVR-L cannot obtain with small residual norm. taking resulted SVR-L, putting initial guess, will give us better solution. To test our SLEs solutions, regular and compared SVR. Numerical results are presented between these two predictors. Lastly, existing interpolation extrapolation methods SLEs. showed performed regression.