1: \begin{abstract}

2:  We propose an extension of quasi-Newton methods, and investigate the convergence and the

3:  robustness properties of the proposed update formulae for the approximate Hessian

4:  matrix. 

5:  Fletcher has studied a variational problem which derives the approximate Hessian update

6:  formula of the quasi-Newton methods. 

7:  We point out that the variational problem is identical to optimization of the

8:  Kullback-Leibler divergence, which is a discrepancy measure 

9:  between two probability distributions. 

10:  Then, we introduce the Bregman divergence as an extension of the Kullback-Leibler

11:  divergence, and derive extended quasi-Newton update formulae based on the variational

12:  problem with the Bregman divergence. 

13:  The proposed update formulae belong to a class of self-scaling quasi-Newton methods. 

14:  We study the convergence property of the proposed quasi-Newton method, and 

15:  moreover, we apply the tools in the robust statistics to analyze the robustness

16:  property of the Hessian update formulae against the numerical rounding errors included in

17:  the line search for the step length. 

18:  As the result, we found that the influence of the inexact line search is bounded only for

19:  the standard BFGS formula for the Hessian approximation. 

20:  Numerical studies are conducted to verify the usefulness of the tools borrowed from

21:  robust statistics. 

22: \end{abstract}

23: