topic Re: Second solve in Proc Optmodel with different starting point (init) in Mathematical Optimization, Discrete-Event Simulation, and OR

Second solve in Proc Optmodel with different starting point (init)

ODENWALD — Wed, 23 May 2018 06:37:56 GMT

After a first solve with NLP with ms-feature I wanted to place a second solve with logic

x.init = x.msinit etc for all variables and solve ..... .

That's not allowed.

Who knows how to implement it ?

Werner

Re: Second solve in Proc Optmodel with different starting point (init)

RobPratt — Wed, 23 May 2018 20:02:19 GMT

The .INIT suffix is deprecated, but you can do the following to set all variables in one loop:

   for {j in 1.._NVAR_} _VAR_[j] = _VAR_[j].msinit;

Re: Second solve in Proc Optmodel with different starting point (init)

ODENWALD — Thu, 24 May 2018 07:35:02 GMT

Thanks, Rob .

.... so easy, but I was looking the wrong direction. Using the ms feature .msinit is clearly more important. But why had you decided to deprecate .init ?

Odenwald

Re: Second solve in Proc Optmodel with different starting point (init)

RobPratt — Thu, 24 May 2018 17:08:29 GMT

The .INIT suffix was confusing (especially because it was not populated by the INIT option in the VAR statement) and not very useful, so we removed it as of SAS/OR 13.2. To supply initial values to be used by the NLP solver, just assign those values to the variables as I showed.