[AMPL 15241] BARON solver converging criteria

classic Classic list List threaded Threaded
3 messages Options
SG
Reply | Threaded
Open this post in threaded view
|

[AMPL 15241] BARON solver converging criteria

SG
Hi,

I tried to use BARON solver but I got the error message:

CPU max time

Then I changed the option and MaxTime=-1

This time, after 8 hours, I got nothing, I wanted to get the global solution but I found only local optimum using filter solver.

That's a real challenge, why BARON cannot be converged!

--
You received this message because you are subscribed to the Google Groups "AMPL Modeling Language" group.
To unsubscribe from this group and stop receiving emails from it, send an email to [hidden email].
To post to this group, send email to [hidden email].
Visit this group at https://groups.google.com/group/ampl.
For more options, visit https://groups.google.com/d/optout.

image.png (606K) Download Attachment
Reply | Threaded
Open this post in threaded view
|

RE: [AMPL 15250] BARON solver converging criteria

Robert Fourer-2
Finding a global optimum is much harder than finding a local one.  Thus sometimes a solver such as BARON is unable to find a provably global optimum in a reasonable amount of time, even though a local solver converges quickly.

You can add outlev=1 to your baron_options string to see the progress that BARON is making.  The listing gives a lower bound and an upper bound on the optimal value of the objective; for a minimization the upper bound is the best solution found so far, and for a maximization the lower bound is the best solution found so far.

Bob Fourer
[hidden email]


-----Original Message-----
From: [hidden email] [mailto:[hidden email]] On Behalf Of SG
Sent: Wednesday, November 29, 2017 9:58 AM
To: AMPL Modeling Language
Subject: [AMPL 15241] BARON solver converging criteria

I tried to use BARON solver but I got the error message:

CPU max time

Then I changed the option and MaxTime=-1

This time, after 8 hours, I got nothing, I wanted to get the global solution but I found only local optimum using filter solver.

That's a real challenge, why BARON cannot be converged!


--
You received this message because you are subscribed to the Google Groups "AMPL Modeling Language" group.
To unsubscribe from this group and stop receiving emails from it, send an email to [hidden email].
To post to this group, send email to [hidden email].
Visit this group at https://groups.google.com/group/ampl.
For more options, visit https://groups.google.com/d/optout.
SG
Reply | Threaded
Open this post in threaded view
|

Re: [AMPL 15253] BARON solver converging criteria

SG
Dear Robert,

I added the new option to show details as:

===========================================================================
  Iteration    Open nodes         Time (s)    Lower bound      Upper bound
          1             1             3.12     1.50000         0.100000E+52
        415           263            48.48     1.50000         0.100000E+52
       1224           641            78.51     1.50000         0.100000E+52
       1417           741           108.54     1.50000         0.100000E+52
       2233          1183           138.58     1.50000         0.100000E+52
       3046          1669           168.61     1.50000         0.100000E+52
       3887          2171           198.67     1.50000         0.100000E+52
       4691          2622           228.71     1.50000         0.100000E+52
       5532          3054           258.75     1.50000         0.100000E+52
       6252          3428           288.78     1.50000         0.100000E+52
       6805          3708           332.38     1.50000         0.100000E+52
       6965          3789           362.44     1.50000         0.100000E+52
       7554          4060           407.30     1.50000         0.100000E+52
       8370          4414           437.34     1.50000         0.100000E+52
       9159          4834           467.41     1.50000         0.100000E+52
       9470          5001           500.40     1.50000          60.0000    

                    *** Max. allowable time exceeded ***      

                   *** No feasible solution was found ***

 Wall clock time:                   500.00
 Total CPU time used:               500.41

 Total no. of BaR iterations:    9470
 Best solution found at node:      -3
 Max. no. of nodes in memory:    5001
 
 All done
===========================================================================
BARON 16.7.29 (2016.07.29): 9470 iterations, CPU time limit reached.
Objective 1e+51
z = 0

I defined the objetive function minimize 

z: 10e4*(X_Am_c + X_Am_g + X_Am_k);

while

# area
param Am_min := 0.5; # min  area [m2]
param Am_max := 20; # max  area [m2]
var X_Am_c >= Am_min, <= Am_max; #  area
var X_Am_g >= Am_min, <= Am_max; #  area
var X_Am_k >= Am_min, <= Am_max; #  area

When I run filter solver I get the optimal value of 

filterSQP (20020316): Optimal solution found, objective = 10.94671379
37 iterations (19 for feasibility)
Evals: obj = 26, constr = 44, grad = 27, Hes = 32
z = 10.9467
X_Am_c = 1.08342
X_Am_g = 5.00282
X_Am_k = 4.86047

I am wondering if I can improve the rate of convergence? Or somehow trigger the solver to see the global optimum 

Regards,





On Wednesday, November 29, 2017 at 9:05:32 PM UTC-5, Robert Fourer wrote:
Finding a global optimum is much harder than finding a local one.  Thus sometimes a solver such as BARON is unable to find a provably global optimum in a reasonable amount of time, even though a local solver converges quickly.

You can add outlev=1 to your baron_options string to see the progress that BARON is making.  The listing gives a lower bound and an upper bound on the optimal value of the objective; for a minimization the upper bound is the best solution found so far, and for a maximization the lower bound is the best solution found so far.

Bob Fourer
<a href="javascript:" target="_blank" gdf-obfuscated-mailto="qwh_YsSGAQAJ" rel="nofollow" onmousedown="this.href=&#39;javascript:&#39;;return true;" onclick="this.href=&#39;javascript:&#39;;return true;">am...@...


-----Original Message-----
From: <a href="javascript:" target="_blank" gdf-obfuscated-mailto="qwh_YsSGAQAJ" rel="nofollow" onmousedown="this.href=&#39;javascript:&#39;;return true;" onclick="this.href=&#39;javascript:&#39;;return true;">am...@... [mailto:<a href="javascript:" target="_blank" gdf-obfuscated-mailto="qwh_YsSGAQAJ" rel="nofollow" onmousedown="this.href=&#39;javascript:&#39;;return true;" onclick="this.href=&#39;javascript:&#39;;return true;">am...@...] On Behalf Of SG
Sent: Wednesday, November 29, 2017 9:58 AM
To: AMPL Modeling Language
Subject: [AMPL 15241] BARON solver converging criteria

I tried to use BARON solver but I got the error message:

CPU max time

Then I changed the option and MaxTime=-1

This time, after 8 hours, I got nothing, I wanted to get the global solution but I found only local optimum using filter solver.

That's a real challenge, why BARON cannot be converged!


--
You received this message because you are subscribed to the Google Groups "AMPL Modeling Language" group.
To unsubscribe from this group and stop receiving emails from it, send an email to [hidden email].
To post to this group, send email to [hidden email].
Visit this group at https://groups.google.com/group/ampl.
For more options, visit https://groups.google.com/d/optout.