A new approach to multivariate adaptive regression splines by using Tikhonov regularization and continuous optimization (Q621735): Difference between revisions

Multivariate approximation and data mining using optimization and regularization methods are discussed in this article. High dimensional approximation problems are solved by a Tikhonov regularization ansatz with a penalized residual sum of squares. The ensuing problem is then solved by modern continuous optimization methods such as interior point methods. At first, the multivariate adaptive regression spline method is introduced and next the aforementioned ansatz with the penalized residual sum of squares used. The optimization methods are then discussed and numerical examples are given.