Topic
  • 3 replies
  • Latest Post - ‏2014-04-28T10:02:07Z by davidoff
MFC
MFC
5 Posts

Pinned topic How to control optimization parameters while using multi-objective feature (staticLex)

‏2013-06-23T17:45:39Z |

Hi, 

I have a question about setting the optimization parameters such as FailLimit, TimeLimit, or OptimalityTolerance while using staticLex (multi objective) functionality. Let's say  we have two objectives staticLex(obj1, obj2). How can we control parameters for first and second objectives? When we set lets say FailLimit=1000, will it effect the first objective or is it also applied to the second objective?

Thanks a lot in advance. 

  • PhilippeLaborie
    PhilippeLaborie
    47 Posts
    ACCEPTED ANSWER

    Re: How to control optimization parameters while using multi-objective feature (staticLex)

    ‏2013-07-01T07:57:10Z  

    Hello,

    Limits and tolerances apply to the search globally, even when you are using staticLex as objective in the model. If you want to have a more control on the resolution of multi-objective problems, you can solve the problem in several steps, each step focusing on some part of the objective. For doing that you may consider using the warmstart functionality of CP Optimizer. With warmstart, you can start the search from a previously known solution. See the example sched_goalprog in the delivery of CP Optimizer.

    Philippe

  • PhilippeLaborie
    PhilippeLaborie
    47 Posts

    Re: How to control optimization parameters while using multi-objective feature (staticLex)

    ‏2013-07-01T07:57:10Z  

    Hello,

    Limits and tolerances apply to the search globally, even when you are using staticLex as objective in the model. If you want to have a more control on the resolution of multi-objective problems, you can solve the problem in several steps, each step focusing on some part of the objective. For doing that you may consider using the warmstart functionality of CP Optimizer. With warmstart, you can start the search from a previously known solution. See the example sched_goalprog in the delivery of CP Optimizer.

    Philippe

  • davidoff
    davidoff
    51 Posts

    Re: How to control optimization parameters while using multi-objective feature (staticLex)

    ‏2014-04-28T09:58:56Z  

    Hello,

    Limits and tolerances apply to the search globally, even when you are using staticLex as objective in the model. If you want to have a more control on the resolution of multi-objective problems, you can solve the problem in several steps, each step focusing on some part of the objective. For doing that you may consider using the warmstart functionality of CP Optimizer. With warmstart, you can start the search from a previously known solution. See the example sched_goalprog in the delivery of CP Optimizer.

    Philippe

    Hello

    I'm continuing this question since it is very interesting from a user prospective to use static Lex directly instead of writing a more complex scheme as Philippe suggests.

    In other frameworks, you can simply set time (or other criterion) limits for the first objective , and an other one for the second.

    If the optimality of the first criterion is not proven, will CPO takes all the available time to improve it regardless of the second criterion ?

    Thanks

    David

  • davidoff
    davidoff
    51 Posts

    Re: How to control optimization parameters while using multi-objective feature (staticLex)

    ‏2014-04-28T10:02:07Z  
    • davidoff
    • ‏2014-04-28T09:58:56Z

    Hello

    I'm continuing this question since it is very interesting from a user prospective to use static Lex directly instead of writing a more complex scheme as Philippe suggests.

    In other frameworks, you can simply set time (or other criterion) limits for the first objective , and an other one for the second.

    If the optimality of the first criterion is not proven, will CPO takes all the available time to improve it regardless of the second criterion ?

    Thanks

    David

    Well apparently the other related post

    https://www.ibm.com/developerworks/community/forums/html/topic?id=77777777-0000-0000-0000-000014927470&ps=25

    explains that staticLex is not a 2 step optimization scheme.