[Ipopt-tickets] [Ipopt] #100: cg-penalty crash / questions

[Ipopt]

#100: cg-penalty crash / questions
------------------------------+---------------------------------------------
Reporter:  hmartin            |       Owner:  ipopt-team
    Type:  defect             |      Status:  new       
Priority:  normal             |   Component:  Ipopt     
 Version:  3.6 (C++ Version)  |    Severity:  normal    
Keywords:  cg-penalty         |  
------------------------------+---------------------------------------------
 Hi,

 After reading some papers of Chen/Goldfarb about their piecewise linear
 penalty method that has been implemented in Ipopt, I decided to give it a
 try.

 However, I got some crashes that I could solve by doing 3 code
 modifications. It seems to me that these are actual bugs (at least for the
 first two, the last modification I did is just a hack). I join my modified
 file hoping it can help.

 By the way, what is the status of this option? [[BR]]
 There is no paper on the Ipopt website describing the implementation, and
 the code is not always documented. Should we expect this option to perform
 in general better than the standard filter line search? [[BR]]
 Is it compatible with other nice options (like the adaptive mu update)?
 [[BR]]
 Reading the available papers, it seems to me this option should perform
 better than the standard Ipopt line search in case of degenerate
 constraints. Does this apply to complementarity constraints (if
 reformulated with a NCP function to satisfy the MFCQ)?
 [[BR]]

 Best regards

-- 
Ticket URL: <https://projects.coin-or.org/Ipopt/ticket/100>
Ipopt <http://projects.coin-or.org/Ipopt>
Interior-point optimizer for nonlinear programs.