[Ipopt] Dual infeasible ?

Rishi Amrit rishiamrit at gmail.com
Mon Sep 17 07:35:59 EDT 2012


Hi,

  I am trying to run a problem, for which I start ipopt with a feasible
initial guess. I set the following options to ensure that it does not lose
feasibility:

bound_push 1e-16
bound_frac 1e-16

The beginning of IPOPT's output looks like:

Number of nonzeros in equality constraint Jacobian...:        0
Number of nonzeros in inequality constraint Jacobian.:    39201
Number of nonzeros in Lagrangian Hessian.............:        0

Total number of variables............................:     4748
                     variables with only lower bounds:       70
                variables with lower and upper bounds:     2728
                     variables with only upper bounds:        0
Total number of equality constraints.................:        0
Total number of inequality constraints...............:     4678
        inequality constraints with only lower bounds:        0
   inequality constraints with lower and upper bounds:     4678
        inequality constraints with only upper bounds:        0

iter    objective    inf_pr   inf_du lg(mu)  ||d||  lg(rg) alpha_du
alpha_pr  ls
   0  5.6444217e+02 0.00e+00 1.35e+00   0.0 0.00e+00    -  0.00e+00
0.00e+00   0
   1  5.6415974e+02 0.00e+00 3.65e+03  -3.5 5.55e-01    -  1.23e-01
9.04e-03f  1
   2  4.4849516e+02 2.32e-02 5.60e+03  -3.6 1.35e+01    -  8.45e-02
2.09e-01f  1
   3  4.0317649e+02 3.07e-02 5.32e+03  -3.6 2.29e+01    -  7.06e-02
5.55e-02f  1
   4  8.8181412e+01 1.09e-01 2.34e+03  -3.6 2.32e+01    -  5.91e-01
5.62e-01f  1


We see that in iter 0, I have feasibility. But it soon looses that, both my
primal and dual infeasibility go up, eventually ipopt enters restoration
mode and fails.

Could you possibly point me to a reason why am I loosing infeasibility ?
and how do I stop it from doing that?

Thanks,

Kind regards,
Rishi
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/ipopt/attachments/20120917/81556f14/attachment.html>


More information about the Ipopt mailing list