[Ipopt] recalc_y option for limited memory Hessian approximation

Chunhua Men chhmen at gmail.com
Fri Apr 21 15:09:17 EDT 2017


I got he convergence issue by using limited memory Hessian approximation
option for my test case - the dual infeasibility value was very high.  So I
played with recalc_y a little bit. But I got the exactly same solution for
each iteration by setting it as "yes" and as "no". I tried different
optimization models, and I always got the same solutions. By setting it as
"yes", I also changed the "recalc_y_feas_tol" to different values, but the
solution from each iteration still did not change.

What I did to changes the parameters are like these"

Anyone can help me on it? or is there any other settings to make "limited
memory Hessian approximation" work better? For my case, computing Hess is
too expensive.

Thanks, Chunhua
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/ipopt/attachments/20170421/c749b074/attachment.html>

More information about the Ipopt mailing list