Saving GAmultiobj results calculated in the algorithm mid-way, without stopping the code run.

I am running a GAmultiobj optimization code with 2 objective functions and 21 input variables. The code takes a long time to run and crashes in between (due to some bug which I am trying to fix).

I am plotting the Pareto-Front of the 2 objectives with default settings. I am able to see the points on the Pareto-Front which tell me the values of the 2 objectives (fval) if I pause my code (say 20 such points). However, I want to also know the variable (xval) values for those points on the Pareto-Front (20-by-21 array), even when the optimization algorithm has not completed. This way atleast some of the variable values will be available to me even if the code crashes later.

I have kept the 'UseParallel' setting to 'true'. What I have tried so far:

1. Include an 'OutputFcn' command to keep storing the 'Population' and respective 'Scores'. However, those values do not correspond to the points currently being displayed on the Pareto-Front.
2. Tried the solution posted in this thread but I am unable to find the results (xval and fval) which have been calculated so far in any of the debugger files. Is this because I am using Parallel setting which is causing the input space to get split up ?

Any help would be really appreciated!

Answer 1

[Swastik Sarkar]

Hi @Rohit Sachdeva,

I was able to use OutputFcn to display the Pareto-front values fval and corresponding x, achieving this with the following output function:

function [state, options, optchanged] = dispParetoFront(options, state, flag)
    optchanged = false;
    if isequal(flag, 'iter')
       topRankedIdx = state.Rank == 1;
       fval = state.Score(topRankedIdx,:);
       x = state.Population(topRankedIdx,:);
       disp(fval);
       disp(x);
    end
end

In the code above, fval corresponds to the values being plotted. Below is the corresponding plot:

Please refer to the below plot for one such iteration:

Following is the output of the OutputFcn for that iteration:

                             Average            Average
Generation   Func-count    Pareto distance    Pareto spread
    1           100                 1                 1
  175.7636  213.0926
  222.3589  202.6253
  207.0830  206.5507

  Columns 1 through 12

    5.3488    0.6370    0.1491   -0.5133    0.2020   -1.0246    2.5131    0.3995   -0.9634   -5.1355    3.5102    2.6572
   -1.5666    1.8390   -4.4555    0.8624   -4.0777    4.5399    4.5765   -2.3334    1.2334    2.1562    3.3867   -3.2181
    4.9653   -0.0933   -3.4053    3.1967   -0.6689    3.4599   -0.0144   -5.6348   -1.9415   -0.6209    2.1915    0.5244

  Columns 13 through 21

    1.5702   -4.9496   -2.0260   -1.5951   -0.1794   -6.9090   -1.8514   -1.9264    1.9222
    1.6208    1.9597   -0.6533    2.3225    0.6406   -2.3525    7.0044    0.8590    6.0227
   -1.0549    4.7118    0.7702   -0.2536   -0.7465   -4.6771    0.1338    7.3640    2.5598

Hope this works for you.