7
$\begingroup$

I have this optimization code:

vars1 = Array[Subscript[x, #] &, {4}];
ka = {35, 10, 20, 25};
objectiveFunction = Total[ka.vars1^5];
NMinimize[{objectiveFunction/100, 
Apply[And, Thread[GreaterEqual[vars1, 0]]] && Total[vars1] == 100 &&
Element[vars1, Integers]}, Flatten[vars1], MaxIterations -> 200]

How is it possible in MMA to capture the values of variables and also objective function in each iteration and plot them versus iteration number??(Separate plots or all in one plot?)

$\endgroup$
1

1 Answer 1

9
$\begingroup$
vars1 = Array[x, {4}];
ka = {35, 10, 20, 25};
objectiveFunction = ka.vars1^5;
s = {};
Dynamic@If[Length@s > 3, ListLinePlot[Transpose@s, PlotRange -> {{0, 200}, {0, 100}}, 
                                      GridLines -> {{Length@s}, {}}],,]
k[vars1_] := NMinimize[{objectiveFunction, And @@ Thread[GreaterEqual[vars1, 0]] &&  
                        Tr@vars1 == 100 && vars1 ∈ Integers},
                        vars1, MaxIterations -> 200, StepMonitor :> AppendTo[s, vars1]]
k[vars1]

enter image description here

$\endgroup$
8
  • $\begingroup$ Great belisarius!!! that is what I was looking for but is it also possible to have the monitoring plot during iterations at the same time??I mean exactly when we have the iteration data?Or it is different story? $\endgroup$
    – Alex
    Commented Oct 24, 2013 at 22:40
  • $\begingroup$ @Alex See edit, please $\endgroup$ Commented Oct 24, 2013 at 22:53
  • $\begingroup$ one more thing why when I changing the separated variables that you put in Sow[{x[1], x[2], x[3], x[4]}]] to Sow[vars1] it is not returning the x[1] value but the rest?What if I have 100 variables?Can I do that same way? $\endgroup$
    – Alex
    Commented Oct 24, 2013 at 22:55
  • $\begingroup$ Nice +1. Just how I would do it. Why is the vertical plot range going to 100, when the option sets the limit at 40? $\endgroup$
    – Michael E2
    Commented Oct 24, 2013 at 22:55
  • 2
    $\begingroup$ Great dynamic modelling is really great!!! $\endgroup$
    – Alex
    Commented Oct 24, 2013 at 22:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.