[Coin-ipopt] Hessian Vector product

[Rohallah Tavakoli]

Dear all,

I use IPOPT in a PDE constrained optimization problem (distributed parameter
identification), i like to use second order information instead of hessian
approximation, but direct construction of Hessian is very expensive, e.g.
for 10^5 design variable, we should solve 10^5 additional PDE ! but
computation of Hessian vector product is cheap, needs only solution of two
additional PDE.

I would like to know that how is difficulty of implementation of this issue
in IPOPT and what portion of code should be modified?

Also, is this issue in plan of developers (i think it considerably increase
capability and range of application of IPOPT)?

Thanks.

Best Regards,
RT.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.coin-or.org/pipermail/ipopt/attachments/20071028/cfc13b46/attachment.html