<meta http-equiv="Content-Type" content="text/html; charset=utf-8" /><P>Hi,</P>
<P>Firstly,thank you for your interest in this E-mail.On the website <A href="http://www.coin-or.org/Ipopt/documentation/node105.html">http://www.coin-or.org/Ipopt/documentation/node105.html</A> presents how to reset option "mehrotra_algorithm".</P>
<P>If option "mehrotra_algorithm" sets to "yes",Ipopt runs as Mehrotra's predictor-corrector </P>
<P>algorithm.This automatically disables the line search, and chooses the (unglobalized) adaptive </P>
<P>mu strategy with the "probing" oracle, and uses "corrector_type=affine" without any safeguards;</P>
<P>you should not set any of those options explicitly in addition.So in my opinion,only reseting </P>
<P>option "mehrotra_algorithm yes" equals to reseting options "mu_strategy adaptive"+"mu_oracle probing"</P>
<P>+"corrector_type affine",but different iteration processing appeared when solving nonlinear optimization</P>
<P>problems under aforementioned two configuration,and the former manifests worse results.Can someone </P>
<P>explain this phenomenon for me? </P>
<P>The second question: In the paper "Adaptive Barrier Strategies for Nonlinear Interior Methods" part 7 </P>
<P>"Corrector Steps" presents Mehrotra's predictor-corrector(MPC) algorithm in detail,points out the complete</P>
<P>MPC step is the sum of the primal-dual and corrector steps. Here,in IPOPT,keyword "corrector_type" has three</P>
<P>options "none,affine,primal-dual",what's the difference between affine and primal-dual,in that paper only </P>
<P>presents one type of method to calculate corrector step.</P>
<P>The third question: Also in that paper part 5 "A Globalization Method".As long as the adaptive primal-dual </P>
<P>steps make sufficient progress towards the solution, the algorithm is free to choose a new value for the </P>
<P>barrier parameter at every iteration. We call this the free mode. However, if the iteration fails to maintain </P>
<P>progress, then the algorithm reverts to a monotone mode, in which a Fiacco-McCormick strategy is applied. </P>
<P>Here, the value of the barrier parameter remains fixed, and a robust globalization technique (e.g., based on<BR>a merit function or a filter) is employed to ensure progress for the corresponding barrier problem. Once the </P>
<P>barrier problem is approximately minimized, the barrier parameter is decreased. The monotone mode continues </P>
<P>until an iterate is generated that makes sufficient progress for the original problem, at which point the free mode resumes. Does this globalization procedure apply in current version of IPOPT?</P>
<P>Sincerely,From Wei.</P>