bfgs_upd.m

Performs a one-step BFGS Hessian update for maximisation using the argument and gradient increments from the previous step.

Syntax

H=bfgs_upd(H,dx,dg)

Parameters

H      - current BFGS approximation to the Hessian
            matrix corresponding to the *negative*
            Hessian of the objective; use [] on the
            first call

   dx     - increment in arguments between the current
            and the previous step

   dg     - increment in gradients between the current
            and the previous step

Returns:

   H      - updated BFGS approximation to the Hessian
            matrix corresponding to the *negative*
            Hessian of the objective

See also

fmaxnewton.m, lbfgs.m, hessreg.m, bracketing.m, sectioning.m, Optimal_control_module

Version 2.11, authors: Ilya Kuprov