[Ipopt] Mehrotra's predictor-corrector algorithm

[zjuzw at sohu.com]

Hi,
Firstly,thank you for your attention to this E-mail.On the website 
http://www.coin-or.org/Ipopt/documentation/node105.html presents how to reset the keyword 
"mehrotra_algorithm". If set to yes, Ipopt runs as Mehrotra's predictor-corrector algorithm. 
This automatically disables the line search, and chooses the adaptive mu strategy with the 
"probing" oracle, and uses "corrector_type=affine" without any safeguards,you should not set 
any of those options explicitly in addition.In my opinion,resetting option "mehrotra_algorithm yes"
equals to resetting options "mu_strategy adaptive" + "mu_oracle probing" + "corrector_type affine",
however the different iteration processing appeared when solving nonlinear optimization problems under
aforementioned two configurations,the former configuration automatically disables the line search and 
manifests worse performance,and the latter configuration still enables the line search,such as resetting
option max_soc nonzero number.Can someone explain this phenomenon for me? Which is the correct configuraton
that brings Mehrotra's predictor-corrector algorithm into effect? 
There are three options of keyword "corrector_type":none,affine,primal-dual.What's the difference between
affine corrector and primal-dual corrector? In the paper "Adaptive Barrier Strategies for Nonlinear Interior 
Methods" can be found on website http://epubs.siam.org/doi/pdf/10.1137/060649513 ,part 7 "Corrector Steps" 
only presents one method to calculate corrector steps.
Sincerely.
From Wei
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/ipopt/attachments/20130322/92180f31/attachment.html>