Decision Optimization

Decision Optimization

Delivers prescriptive analytics capabilities and decision intelligence to improve decision-making.


#Analytics
#DecisionOptimization
#DecisionOptimization
 View Only
  • 1.  LP is solved but crashed at the root node

    Posted 10/17/14 07:04 AM

    Originally posted by: Falcon_G


    Hi everybody

    I have a nasty MIP for which the LP is solved in two hours.

    The MIP crashes at the root node in less than 100 seconds .

    I have difficulties to understand this. How comes that CPLEX crashes at the root node even before the LP being solved.

    I know that this crash is somehow the limit of a 32-bit OS imposed on CPLEX.

    I read somewhere in CPLEX forum that for obtaining LP obj we set nodelim to 0 and get objective value and set nodeim to maximum and continue MIP optimization.

    I am also not sure if this is correct.

    My second question is if I run CPLEX to solve a MIP, is there any correct way to record the LP relaxation objective value without need to solve LP separately by relaxing integratlity on variables.

     

    Any constructive comment is appreciated.


    #CPLEXOptimizers
    #DecisionOptimization


  • 2.  Re: LP is solved but crashed at the root node

    Posted 10/22/14 03:25 AM

    If the LP solves without problem but MIP runs out of memory then the first thing to try is to set either CPX_PARAM_THREADS to 1 or CPX_PARAM_AUXROOTTHREADS to -1. Additional threads may require additional memory (in particular additional copies of the model) which may increase the memory consumption compared to pure LP solves.

    You can track the objective function of the relaxation by using a callback (a solve callback or a branch callback, for example). This is discussed in several threads on this forum. If you only want the objective value at the root node then you can indeed set the node limit to 0 and query the dual bound after solve() returns. That should be the objective function value of the LP relaxation (including cuts).


    #CPLEXOptimizers
    #DecisionOptimization


  • 3.  Re: LP is solved but crashed at the root node

    Posted 10/22/14 03:44 AM

    Originally posted by: Falcon_G


    Thank you indeed.

    Regarding the second question: "That should be the objective function value of the LP relaxation (including cuts). "

    But if that is "including cuts" is should be something stronger than LP bound unless all those cuts are redundant? isn't it?will it be still the LP bound?

     

     


    #CPLEXOptimizers
    #DecisionOptimization


  • 4.  Re: LP is solved but crashed at the root node

    Posted 10/28/14 02:53 AM

    Yes, the dual bound at the end of the root includes cuts and should thus be stronger than the bound provided by the initial relaxation. If you want to track the dual bound of the initial relaxation then you have to use a callback.


    #CPLEXOptimizers
    #DecisionOptimization