1: \begin{abstract}

2: An outstanding problem when computing a function of a matrix, $f(A)$,

3: by using a Krylov method is to accurately estimate errors when

4: convergence is slow.  Apart from the case of the exponential function

5: which has been extensively studied in the past, there are no

6: well-established solutions to the problem.  Often the quantity of

7: interest in applications is not the matrix $f(A)$ itself, but rather,

8: matrix-vector products or bilinear forms. When the computation related

9: to $f(A)$ is a building block of a larger problem (e.g., approximately

10: computing its trace), a consequence of the lack of reliable error

11: estimates is that the accuracy of the computed result is unknown. In

12: this paper, we consider the problem of computing $\tr(f(A))$ for a

13: symmetric positive-definite matrix $A$ by using the Lanczos method and

14: make two contributions: (i) we propose an error estimate for the

15: bilinear form associated with $f(A)$, and (ii) an error estimate for

16: the trace of $f(A)$.  We demonstrate the practical usefulness of these

17: estimates for large matrices and in particular, show that the trace

18: error estimate is indicative of the number of accurate digits. As an

19: application, we compute the log-determinant of a covariance matrix in

20: Gaussian process analysis and underline the importance of error

21: tolerance as a stopping criterion, as a means of bounding the number

22: of Lanczos steps to achieve a desired accuracy.

23: \end{abstract}

24: