1Blurb::
2Hessians are needed and will be approximated by secant updates (BFGS
3or SR1) from a series of gradient evaluations
4
5Description::
6The \c quasi_hessians specification means that Hessian information is
7needed and will be approximated using secant updates (sometimes called
8"quasi-Newton updates", though any algorithm that approximates
9Newton's method is a quasi-Newton method).
10
11Compared to finite difference numerical Hessians, secant
12approximations do not expend additional function evaluations in
13estimating all of the second-order information for every point of
14interest. Rather, they accumulate approximate curvature information
15over time using the existing gradient evaluations.
16
17The supported secant approximations include the
18Broyden-Fletcher-Goldfarb-Shanno (BFGS) update (specified with the
19keyword \c bfgs) and the Symmetric Rank 1 (SR1) update (specified with
20the keyword \c sr1).
21
22Topics::
23Examples::
24Theory::
25
26
27
28Faq::
29See_Also::	responses-no_hessians, responses-numerical_hessians, responses-analytic_hessians, responses-mixed_hessians
30