GMRES with adaptively deflated restarting and its performance on an electromagnetic cavity problem (Q550089)

The paper focuses on the generalized minimal residual (GMRES) method with deflated restarting, denotied by GMRES-DR\((m,k)\), for solving nonsymmetric linear systems. The author presents a new deflation strategy, called GMRES-ADR, which determines \(k\) dynamically in different cycles. The first section is an introduction in nature. The second section briefly presents GMRES-DR\((m,k).\) In the third section, assuming exact precision arithmetic, the author gives some theoretical analysis of GMRES-DR\((m,k).\) In particular, he shows that the convergence rate of one cycle of GMRES-DR\((m,k)\) is similar to the one obtained by removing the components of the initial residual vector associated with the deflated eigenvalues. The fourth section proposes the deflation strategy used in the numerical experiments. In the fifth section one reports numerical results of GMRES-ADR for an electromagnetic cavity problem and the main observations concerning the efficiency of this new deflation strategy. The last section presents some concluding remarks and topics for future work.