Based on the generalised Arnoldi procedure, we develop an implicitly restarted
generalised Arnoldi method for solving the large-scale polynomial eigenvalue problem.
By combining implicit restarting with the refinement scheme, we present an implicitly
restarted refined generalised Arnoldi (IRGAR) method. To avoid repeated converged
eigenpairs in the later iteration, we develop a novel non-equivalence low-rank deflation
technique and propose a deflated and implicitly restarted refined generalised Arnoldi
method (DIRGAR). Some numerical experiments show that this DIRGAR method is efficient