1: \begin{abstract}

2: A generalized skew-symmetric Lanczos bidiagonalization (GSSLBD) 

3: method is proposed to compute several extreme eigenpairs 

4: of a large matrix pair $(A,B)$, where $A$ is skew-symmetric 

5: and $B$ is symmetric positive definite. 

6: The underlying GSSLBD process produces two sets of $B$-orthonormal 

7: generalized Lanczos basis vectors that are also $B$-biorthogonal 

8: and a series of bidiagonal matrices whose singular values are 

9: taken as the approximations to the imaginary parts of the 

10: eigenvalues of $(A,B)$ and the corresponding left and right 

11: singular vectors premultiplied with the left and right generalized 

12: Lanczos basis matrices form the real and imaginary parts of the 

13: associated approximate eigenvectors.  

14: A rigorous convergence analysis is made on the desired eigenspaces 

15: approaching the Krylov subspaces generated by the GSSLBD process 

16: and accuracy estimates are made for the approximate eigenpairs. 

17: In finite precision arithmetic, it is shown that the 

18: semi-$B$-orthogonality and semi-$B$-biorthogonality 

19: of the computed left and right generalized Lanczos vectors 

20: suffice to compute the eigenvalues accurately. 

21: An efficient partial reorthogonalization strategy is adapted 

22: to GSSLBD in order to maintain the desired semi-$B$-orthogonality and 

23: semi-$B$-biorthogonality. 

24: To be practical, an implicitly restarted GSSLBD algorithm, abbreviated 

25: as IRGSSLBD, is developed with partial $B$-reorthogonalizations.

26: Numerical experiments illustrate the robustness and overall 

27: efficiency of the IRGSSLBD algorithm.

28: \end{abstract}

29: