[Ipopt] Request for Information for any IPOPT termination hacks for Convex functions

[jmogali at andrew.cmu.edu]

Hi,
     I have a query about the termination criterion for IPOPT. So the
function I am trying to minimize is a convex self concordant function
subjected to linear constraints. I observed that a large number of
iterations have nearly identical objective, inf_pr,inf_du
lg(mu),||d||,lg(rg),alpha_du, alpha_pr with 1 as ls (backtracking
line search parameter). Since my problem is convex self concordant,
is there a way I can reduce the number of
iterations ? Right now many iterations (~300) just keep happening with
almost the same values.

I modified "tol" and "dual_inf_tol", however since my problem is convex
the default "dual_inf_tol" of 1 is already large, so I did not modify that
value, the rest of them are set to the default values. Do you recommend
changing any other parameter ?

In general, I would like to get recommendations on using IPOPT for convex
functions such that I can exploit the convexity fact to reduce the number
of IPOPT iterations before termination.

Thanks,
Jayanth