Multidirectional subspace expansion for one-parameter and multiparameter Tikhonov regularization (Q2356601)

The authors present a new method for large scale Tikhonov regularization problems which combines a multidirectional subspace expansion with optional truncation to produce a higher quality search space. The multidirectional expansion generates a richer search space whereas the truncation ensures moderate growth. The authors present lower and upper bounds on the regularization parameter when the discrepancy principle is applied to one-parameter regularization, and use numerical results to illustrate that their method can yield more accurate results or faster convergence.