How to get all the solutions in detail from a solver NLP were there is many distinct local optima?

POPO1989 — Sun, 05 Jan 2020 11:59:43 GMT

Hi, I'm new on SAS 9. 4.

I have an objective to minimize:

min F = qte[17]/71 + qte[13]/25 + (qte[31] + qte[32])/263 - qte[1]/272 - qte[4]/254 - qte[6]/166 - qte[10]/268;

(where qte is a vector in which each value is the % recommanded to eat for a nutrient in a diet, and the aim of my program is to find those %)

I choose this way of solving:

solve obj F with NLP/algorithm=Activeset /*Concurrent*/ multistart Maxiter=10000 MSMaxstarts=25 seed=1863071 Feastol=0.001;

I get on the "journal" (I have SAS in french) the following message:

NOTE: Problem generation will use 4 threads.
NOTE: The problem has 36 variables (0 free, 0 fixed).
NOTE: The problem has 34 linear constraints (23 LE, 0 EQ, 0 GE, 11 range).
NOTE: The problem has 1025 linear constraint coefficients.
NOTE: The problem has 0 nonlinear constraints (0 LE, 0 EQ, 0 GE, 0 range).
NOTE: The OPTMODEL presolver removed 0 variables, 3 linear constraints, and 0 nonlinear constraints.
NOTE: The OPTMODEL presolved problem has 36 variables, 31 linear constraints, and 0 nonlinear constraints.
NOTE: The OPTMODEL presolver removed 36 linear constraint coefficients, leaving 989.
NOTE: Using analytic derivatives for objective.
NOTE: The NLP solver is called.
NOTE: The Active Set algorithm is used.
NOTE: The MULTISTART option is enabled.
NOTE: The deterministic parallel mode is enabled.
NOTE: The Multistart algorithm is executing in single-machine mode.
NOTE: The Multistart algorithm is using up to 4 threads.
NOTE: Random number seed 1863071 is used.
      Meilleur       Local  Optimalit?    Infaisa-  Local  Local
      D?but    Objectif   Objectif       Erreur      bilit?  It?r  Statut
          1   -5,7473197  -5,7473197  8,19617E-7  4,5475E-13    165  Optimal
          2   -5,7473197  -4,4371758  7,44232E-7   9,9268E-6    150  Optimal
          3 * -5,7473197   -4,193588  7,58871E-7  0,00001819    188  Optimal
          4   -5,7473197  -2,4594022  9,88805E-7   0,0000401    220  Optimal
          5   -5,7473197  -3,4290596  9,23225E-7  8,75817E-7    143  Optimal
          6   -6,1896441  -6,1896441  1,51637E-7  0,00098512    158  Optimal
          7   -6,1896441  -13562,005  5,37896E-7   100460631     76  LocalInfeas
          8   -6,1896441  -4,5812214  9,27801E-7  0,00002634    270  Optimal
          9   -6,1896441  -4,3003702  9,53107E-7           0    395  Optimal
         10   -6,1896441  -13566,637  7,45639E-7  90114971,5    110  LocalInfeas
         11   -6,1896441  -4,3436519  9,56977E-7  3,07991E-7    577  Optimal
         12   -6,1896441  -13570,349  5,22808E-6   111100464  10000  IterLimit
         13   -6,1896441  -1,2855454  8,22242E-6  52,1652856  10000  IterLimit
         14   -6,1896441  3802,28137   0,0000384  5782580,77  10000  IterLimit
         15   -6,1896441  -1,2521854   0,0000384  217,518794  10000  IterLimit
         16   -6,1896441  -9624,2172  6,79356E-7   148508958     57  LocalInfeas
         17   -6,1896441  -13565,637  6,58687E-7   156238573     66  LocalInfeas
         18   -6,1896441  -9629,6289  6,81325E-7  86409963,5     95  LocalInfeas
         19   -6,1896441  -13565,637  0,00003841   507821170  10000  IterLimit
         20   -6,1896441  -7541,5404  9,90461E-7  90599996,7     58  LocalInfeas
         21   -6,1896441  -0,0650454  1,83072E-6  1323,91452  10000  IterLimit
         22   -6,1896441  -9762,3554  5,40408E-8  96672409,1     56  LocalInfeas
         23   -6,1896441  -6149,1399  0,00002531  11529782,4  10000  IterLimit
         24   -6,1896441  -8605,4825  2,93296E-6  92399924,9  10000  IterLimit
         25   -6,1896441  -3914,6655  0,00003845  92303556,9  10000  IterLimit
NOTE: The Multistart algorithm generated 1600 sample points.
NOTE: 9 distinct local optima were found.
NOTE: The best objective value found by local solver = -6,189644118.
NOTE: The solution found by local solver with objective = -6,189644118 was returned.

I don't understand all the content, but,

As I would like to see the details not only of the best objective, but also for all the other distinct local optima, I tried this code:

print _NSOL_;
print {j in 1..36, i in 1.._NSOL_} qte[j].sol[i];
create data solout from [sol]={i in 1.._NSOL_}
{j in 1..36} <col("qte"||j)=qte[j].sol[i]> ;


print F; 
create data INCA3_9C.regime_opt from [i] quantite=qte;
quit;

proc print data= solout;
run;

but it didn't work.

If anybody have a solution for my problem, please, let me know, this will help me a lot !

Re: How to get all the solutions in detail from a solver NLP were there is many distinct local optim

Ksharp — Mon, 06 Jan 2020 11:10:35 GMT

Post it at OR forum and calling @RobPratt

Re: How to get all the solutions in detail from a solver NLP were there is many distinct local optim

RobPratt — Mon, 06 Jan 2020 14:56:16 GMT

Only the best solution among these distinct local optima is returned, and so _NSOL_ = 1. Currently, there is no way to access the other solutions.

topic Re: How to get all the solutions in detail from a solver NLP were there is many distinct local optim in New SAS User

How to get all the solutions in detail from a solver NLP were there is many distinct local optima?

Re: How to get all the solutions in detail from a solver NLP were there is many distinct local optim

Re: How to get all the solutions in detail from a solver NLP were there is many distinct local optim