1: \begin{abstract}

2: Solving the trust-region subproblem (TRS) plays a key role in

3: numerical optimization and many other applications.

4: The generalized Lanczos trust-region (GLTR) method is a well-known Lanczos type

5: approach for solving a large-scale TRS.

6: The method projects the original large-scale TRS

7: onto a $k$ dimensional Krylov subspace,

8: whose orthonormal basis is generated by the symmetric Lanczos process, and computes

9: an approximate solution from the underlying subspace. There have been

10: some a-priori error bounds for the optimal solution and the optimal objective value

11: in the literature, but no a-priori result exists on

12: the convergence of Lagrangian multipliers involved

13: in projected TRS's and the residual norm of approximate solution. In this paper,

14: a general convergence theory of the GLTR method is

15: established, and a-priori bounds are derived for the errors of the

16: optimal Lagrangian multiplier, the optimal solution, the optimal

17: objective value and the residual norm of approximate solution.

18: Numerical experiments demonstrate that our bounds are realistic and

19: predict the convergence rates of the three errors and residual norms accurately.

20: \end{abstract}